Modern financial networks, particularly those supporting High-Frequency Trading (HFT) systems, are characterized by intense data flows and high packet rates, where even minor variations in latency can impact trading outcomes. The problem of queuing and latency management in such networks has become increasingly critical. Dynamic load balancing comes as a solution to this by splitting traffic and sending it on multiple paths. Usually, the problem observed is re-ordering. In response to this challenge, we present a novel Software-Defined Load Balancing Framework for splitting traffic at the packet level without causing packet reordering, which is especially suited for High-Frequency Trading Networks. We propose SOLT+ a traffic splitting algorithm that inspects bursts of packets specifically choosing packets to avoid re-ordering. We perform MATLAB simulations to show the accuracy and compare SOLT + with state-of-the-art algorithms.

Feature-based Visual Simultaneous Localization and Mapping (V-SLAM) employs feature descriptors to recognize landmarks in successive frames. For networked (offloaded and/or collaborative) V-SLAM, the feature descriptors typically need to be transmitted over wireless communication networks with limited bitrates. Conventionally, V-SLAM systems have employed general computer vision feature descriptors whose size (in bytes) accounts for most of the data transmission. We develop a probabilistic model of the V-SLAM feature matching and identify three parameters that govern the probability of the correct feature match: the feature set size, the bit error probability between feature descriptors observing the same landmark, and the probability that the correct feature match exists in the feature set. Based on these three parameters, we formulate a research agenda for achieving high V-SLAM performance (i.e., high correct feature match probability) for reduced V-SLAM feature descriptor sizes (and thus reduced network throughput requirements). We address the first item on this research agenda by pursuing the reduction of the feature set size through a novel Orientation Restriction with Frame-to-Frame Projection (ORFFP). We evaluate ORFFP in comparison to the state-of-the-art ORB-SLAM2 approach using three representative V-SLAM test sequences. Our ORFFP approach reduces the size of the feature set to as little as 9.1% of ORB-SLAM2, allowing for the use of smaller feature descriptors. Thereby, ORFFP reduces the overall required throughput for networked V-SLAM down to roughly a third of the current approach, while achieving essentially the same V-SLAM performance. We also find that for learned descriptor-based models (HashSIFT, BAD), the feature descriptor size can be reduced using ORFFP without severely degrading the performance of V-SLAM. We outline future research directions towards comprehensively addressing the formulated research agenda for enabling networked V-SLAM on low-bitrate networks.

Emerging 5G/6G use cases span various industries, necessitating flexible solutions that leverage emerging technologies to meet diverse and stringent application requirements under changing network conditions. The standard 5G RAN packet error handling using retransmission reduces packet loss but can increase transmission delay. Random Linear Network Coding (RLNC) offers an alternative by proactively sending combinations of original packets, thus reducing both delay and packet loss. Previous research typically only simulates the integration of RLNC in 5G but does not demonstrate nor evaluate this integration in real 5G systems. In contrast, we implement and evaluate our approach through measurements with commercially available servers and switches, running the OpenAirInterface (OAI) 5G software stack. We introduce Flexible Network Coding (FlexNC), which enables the flexible fusion of several RLNC protocols. Specifically, FlexNC provides a forwarder that flexibly interfaces with multiple RLNC protocols in a cloud-native (containerized) solution. Network operators can configure FlexNC based on network conditions and application requirements. For boosting network programmability, our Recoder in the Network (RecNet) leverages In-Network Computing (INC) in intermediate network nodes. We have developed an open-source cloud-native (Docker-based) implementation of both FlexNC and RecNet on OAI, including INC for the Recoder on a Cisco Catalyst switch. Measurements for video, haptic, and audio traffic indicate that i.) FlexNC adapts to various application needs in terms of latency and packet loss, and RecNet significantly reduces packet loss for a remote user with minimal increase in delay compared to pure RLNC. To the best of our knowledge, this is the first article to report testbed measurement results for cloud-native network coding in a 5G system, thus creating a baseline for reliable 5G communication.

Emerging network application paradigms, such as the Tactile Internet, re-emphasize the need for different Quality of Service (QoS) levels. Due to the large packet buffers in the underlying network data plane, Active Queue Management (AQM) is generally required to curtail packet latencies for flows requiring high QoS levels. At the same time, programmable data planes, such as P4, enable packet processing at line-speed, albeit with limited packet processing functionalities. However, the existing AQM mechanisms that support QoS differentiation are too complex to readily run on P4, while the existing AQM mechanisms that run on P4 do generally not support effective QoS differentiation. We address this gap by developing to the best of our knowledge the first AQM mechanism that supports effective QoS differentiation while running on P4. Specifically, we propose SP-CoDel, which combines the well-known Controlled Delay (CoDel) AQM mechanism with Static Priority (SP) scheduling for QoS differentiation. Also, we propose RED-SP-CoDel, which adds a RED AQM component to SP-CoDel so as to make the AQM with priorities essentially parameterless. As a community resource contribution, we substantially extend the existing P4Simulator to the novel fused P4-NS3 Simulator so as to enable the evaluation of packet processing mechanisms through the combined functionalities of P4 device emulation and NS3 packet simulation. Evaluations conducted with the P4 reference switch model in the P4-NS3 Simulator indicate that SP-CoDel and RED-SP-CoDel provide high QoS to high-priority data streams, i.e., significantly reduce latency and packet loss compared to CoDel, while effectively mitigating bufferbloat.

With the emergence of social media and computing, many users have utilized social media platforms (SMPs) in communicating and sharing their interests and preferences. One critical challenge is that social media have been employed for propaganda and influence campaigns for various purposes, such as spreading fake news. SMPs generate vast amounts of data that demand machine learning (ML) capabilities to efficiently learn and infer influence campaigns. This study proposes an ML framework for the thematic campaign classification (TCC), assisting decision-makers in understanding the impacts of social media toward end-users and aiding the mitigation of their side effects. The proposed framework relies on a newly developed characterization that we have termed the campaign network attribute (CNA), which adopts representative network features for the effective TCC by neural networks. The proposed CNA-TCC framework was validated using Twitter and Instagram data sources. The proposed framework can achieve a classification precision in the range of 68%–90% for two campaigns. Also, it can identify a known campaign in generic social media data with a 60% precision. The empirical results indicated high-performance levels of the proposed CNA-TCC framework that can substantially reduce the search spaces for social influence campaigns on specific themes. The proposed CNA-TCC framework has the potential to be applied in real-world SMPs and to process their large-scale data, so as to effectively classify influence campaigns.

This paper studies artificial intelligence (AI) aided communication and computing resource allocation in a vehicular network that supports blockchain-enabled video streaming. Our study aims to improve the operating efficiency and to maximize the transcoding rewards for blockchain based vehicular networks. Our resource allocation policy considers the vehicular mobility, which is modelled with a highly-realistic Semi-Markov renewal process, as well as the real-time video service delay constraints. We propose a multi-timescale actor-critic-reinforcement learning framework to tackle these grand challenges. We also develop a prediction model for the vehicular mobility by using analysis and classical machine learning, which alleviates the heavy signaling and computation overheads due to the vehicular movement. A mobility-aware reward estimation for the large timescale model is then proposed to mitigate the complexity due to the large action space. Finally, numerical results are presented to illustrate the developed theoretical findings in this paper and the significant performance gains due to our proposed multi-timescale framework.

In order to provide consistent low-latency communication network services, Time-Sensitive Networking (TSN) unites a set of standards for time-synchronization, flow control, enhanced reliability, and management. We design the TSN-FlexTest testbed with generic commodity hardware and open-source software components to enable flexible TSN measurements. We have conducted extensive measurements to validate the TSN-FlexTest testbed and to examine TSN characteristics. The measurements provide insights into the effects of TSN configurations, such as increasing the number of synchronization messages for the Precision Time Protocol, indicating that a measurement precision of 30 ns can be achieved. The TSN measurements included extensive evaluations of the Time-Aware Shaper (TAS) for sets of Tactile Internet (TI) packet traffic streams. The measurements elucidate the effects of different scheduling and shaping approaches, while revealing the need for pervasive network control that synchronizes the sending nodes with the network switches. We present the first measurements of distributed TAS with synchronized senders on a commodity hardware testbed, demonstrating the same Quality-of-Service as with dedicated wires for high-priority TI streams despite a 200% over-saturation cross traffic load. The testbed is provided as an open-source project to facilitate future TSN research.

Computing in the Network (COIN) has the potential to reduce the data traffic and thus the end-to-end latencies for data-rich services. Existing COIN studies have neglected the impact of the size of the data unit that the network nodes compute on. However, similar to the impact of the protocol data unit (packet) size in conventional store-and-forward packet-switching networks, the Computing Data Unit (CDU) size is an elementary parameter that strongly influences the COIN dynamics. We model the end-to-end service time consisting of the network transport delays (for data transmission and link propagation), the loading delays of the data into the computing units, and the computing delays in the network nodes. We derive the optimal CDU size that minimizes the end-to-end service time with gradient descent. We evaluate the impact of the CDU sizing on the amount of data transmitted over the network links and the end-to-end service time for computing the convolutional neural network (CNN) based Yoho and a Deep Neural Network (DNN) based Multi-Layer Perceptron (MLP). We distribute the Yoho and MLP neural modules over up to five network nodes. Our emulation evaluations indicate that COIN strongly reduces the amount of network traffic after the first few computing nodes. Also, the CDU size optimization has a strong impact on the end-to-end service time; whereby, CDU sizes that are too small or too large can double the service time. Our emulations validate that our gradient descent minimization correctly identifies the optimal CDU size.

Motivated by the need to provide a precisely determined delay between source and sink nodes in time-sensitive networks, we propose an architecture that provisions near-zero queuing delay in new Quality-of-Service frameworks, e.g., those of$\mathbf{5G}$solutions. To this end, various studies have shown how load balancing can reduce delay. Most of these studies consider$N$parallel processing queues with exponential service rates and Poisson arrivals with mean rate$\lambda$. These queues are handled by a single controller that assigns a new task to the shortest queue. The so-called power-of-d-servers or power-of-d-choices approach was proven to provide necessary delay improvements. In this strategy, the controller allocates the request to the least-loaded server among$d(N), 1\leq d(N)\leq N$randomly selected servers. However, none of these studies have considered realistic scenarios of fractional resource assignment to flow requests. To address this key shortcoming, we make the following contributions: (1) We design a software-defined network (SDN) controller framework called SOLT that considers the keys aspects of available resources in a time-sensitive network (TSN) setting, (2) We prove theoretically, how these bounds can be achieved and show the necessary conditions for achieving asymptotically zero delays in such networks, and (3) Through simulations, we demonstrate the improvements achieved by SOLT in comparison with state-of-the-art algorithms.

Visual Simultaneous Localization and Mapping (SLAM) systems have their numerous applications in robotics and autonomous driving. Computational offloading enables the computationally demanding Visual SLAM systems to run on hardware-constrained mobile devices, such as Unmanned Aerial Vehicles (UAVs) and Automated Guided Vehicles (AGVs). The offloading of SLAM modules to a central server also enables cooperative collaboration between multiple mobile devices. In this paper, we investigate the process of map synchronization between the edge and mobile devices in edge assisted SLAM systems. Due to the condition of uninterrupted execution, it is necessary that the map synchronization can be executed in the running process without losing the localization on the mobile device. Using a state-of-the-art edge assisted SLAM system, we investigate the influences of the movement speed as well as the network latency on the map synchronization and the resulting success ratio of the tracking continuation. By introducing a frame buffer on the mobile device, we have managed to compensate for the negative effect of the synchronization delay and thus increase reliability by up to 37%.

The trade-offs between storage and repair traffic for replacing failed storage nodes with new nodes (newcomers) in data centers with an omniscient controller are well understood. However, in edge storage settings, newcomers are not readily available, necessitating resilient data storage (caching) without newcomers. Edge storage nodes can often communicate via a broadcast wireless medium, which can be exploited to reduce the transmitted repair traffic via network coding. Repairs for resilient distributed caching without newcomers over a broadcast medium with Random Linear Network Coding (RLNC), which does not require an omniscient controller, have not been previously studied. We develop an information-theoretic model to characterize the theoretically achievable trade-offs between stored data and transmitted repair data in the RLNC broadcast setting without newcomers. Specifically, we formulate an Information Flow Graph (FG) model and identify all cuts in the resulting FG. We validate the theoretical FG model with simulations that demonstrate that the practically achievable trade-offs are close to the theoretical trade-offs.

Identification via channels (ID) is a goal-oriented (Post-Shannon) communications paradigm that verifies the matching of message (identity) pairs at source and sink. To date, ID research has focused on the upper bound$\lambda $for the probability of a false-positive (FP) identity match, mainly through ID tagging codes that represent the identities through ID codeword sets consisting of position-tag tuples. We broaden the ID research scope by introducing novel ID performance metrics: the expected FP-error probability$\overline {p_{\mathrm {fp}}}$which considers distance properties of ID codeword sets in conjunction with the probability for selecting ID pairs, the threshold probabilities$p_{\epsilon }$that characterize quantiles of FP-probabilities, and the distance tail uplift ratio DiTUR giving the fraction of ID pairs whose distance is increased above the minimum distance (which corresponds to$\lambda $). We define a No-Code (NC) approach that directly conducts the ID operations with the messages (identities) without any additional coding as a baseline for ID. We investigate a concatenated Reed-Solomon ID code and a Reed-Muller ID code, and find that they do not always yield advantages over using no ID code. We analytically characterize the reduction of error-prone ID pairs through sending multiple tags. Overall, our insights point to investigating the distance distribution of ID codes and to incorporating the ID pair distributions of real ID systems in future ID research.

5G Millimeter Wave (mmWave) technology holds great promise for Connected Autonomous Vehicles (CAVs) due to its ability to achieve data rates in the Gbps range. However, mmWave suffers from a high beamforming overhead and requirement of line of sight (LOS) to maintain a strong connection. For Vehicle-to-Infrastructure (V2I) scenarios, where CAVs connect to roadside units (RSUs), these drawbacks become apparent. Because vehicles are dynamic, there is a large potential for link blockages. These blockages are detrimental to the connected applications running on the vehicle, such as cooperative perception and remote driver takeover. Existing RSU selection schemes base their decisions on signal strength and vehicle trajectory alone, which is not enough to prevent the blockage of links. Many modern CAVs motion planning algorithms routinely use other vehicle’s near-future path plans, either by explicit communication among vehicles, or by prediction. In this paper, we make use of the knowledge of other vehicle’s near future path plans to further improve the RSU association mechanism for CAVs. We solve the RSU association algorithm by converting it to a shortest path problem with the objective to maximize the total communication bandwidth. We evaluate our approach, titled B-AWARE, in simulation using Simulation of Urban Mobility (SUMO) and Digital twin for self-dRiving Intelligent VEhicles (DRIVE) on 12 highway and city street scenarios with varying traffic density and RSU placements. Simulations show B-AWARE results in a 1.05× improvement of the potential datarate in the average case and 1.28× in the best case vs. the state-of-the-art. But more impressively, B-AWARE reduces the time spent with no connection by 42% in the average case and 60% in the best case as compared to the state-of-the-art methods. This is a result of B-AWARE reducing nearly 100% of blockage occurrences.

Industrial Internet of Things (IIoT) applications have diverse network session requirements. Certain critical applications, such as emergency alert relays, as well as industrial floor evacuation and surveillance systems, require fresh updates that can maintain the most recently delivered packets. This requires high reconfigurability to an extent where the system can measure the impact of an event and adapt the network accordingly. Recent research has demonstrated that network failures can undermine the sustainability of Industry 4.0 or Industrial IoT in general. In this paper, we design an intelligent Federated learning based Time-Sensitive Networking (Fed-TSN) controller framework to optimize the failure recovery. In industrial IoT scenarios, such as emergency evacuations on factory floors due to natural disasters, there can be multiple link failures with no disjoint paths which require a sustainable recovery solution. Accordingly, we consider multiple simultaneous link failures, both for networks with and without disjoint paths. The typically probabilistic network failures on a factory floor call for designing a mechanism that can search for routes with minimum joint failure probability (JFP). We formulate the JFP minimization problem as a non-linear integer program. We design a Software Defined Networking (SDN) controller that runs an application to produce near-optimal solutions for providing enhanced sustainability in a wide range of Industry 4.0 scenarios. We employ this non-linear integer program solution as input to our intelligent Fed-TSN fault recovery strategy that predicts the migration location based on the changes in the TSN gate schedule. We conduct simulations to quantify the improvements achieved with Fed-TSN compared to state-of-the-art approaches.

Existing Network Function Virtualization (NFV) service placements that reuse already deployed network functions either reuse an entire Service Function Chain (SFC) or only individual network functions while ignoring the chain configuration cost for configuring the SFC traffic steering and ignoring the reliability of the network functions. Also, the Mobile Edge Cloud (MEC) frameworks that are required to implement an NFV service placement should ideally seamlessly cooperate with the various existing NFV Management and Orchestration (MANO) frameworks. However, the existing MEC frameworks lack multi-MANO support. We formulate the novel Subchain-Aware NFV service Placement (SAP) optimization model that accounts for the configuration cost for stitching together reused network functions to an SFC and strives to reuse existing subchains of consecutive network functions (with already deployed SFC traffic steering), while accounting for the recovery cost of network functions with limited reliability. We develop Tabu-SAP, a Tabu search approach to solve the SAP optimization problem. Furthermore, we introduce the novel Automated Provisioning framework for MEC (APMEC) with open-source OpenStack implementation to enable the deployment of Tabu-SAP in real networks; APMEC supports multiple MANOs through a loose coupling MANO-MEC design. Our Tabu-SAP evaluations indicate an around eightfold increase of the number of supported SFCs compared to the state-of-the-art reuse of individual network functions, while substantially reducing the total cost, which includes the chain configuration cost. Also, for long SFCs of seven or more network functions, the Tabu-SAP total cost is less than 10% higher than the optimal solution (which requires over ten times longer execution time).

Edge Simultaneous Localization and Mapping (SLAM) retains only the tracking on the mobile device, while offloading the compute-intensive local mapping and loop close to edge computing. Existing Edge SLAM approaches incur relatively high delays for offloading, resulting in high failure probabilities, i.e., low reliability, for commonly used public SLAM datasets. We discovered that two parameters which had not previously been studied in detail, namely the number of features and the number of keyframes that are bundled for a local map update, play a critical role in the offloading delay. Also, previous approaches updated the local map in the mobile device in a serial manner, incurring map update latencies. We study the numbers of features and bundled keyframes in detail and we parallelize the local map update. We find that judicious parameter settings, namely relatively small numbers of features (750 per frame) and bundled keyframes (1, i.e., effectively no bundling), reduce the map update latency to less than half compared to the previously common settings (1000 features per frame and 6 keyframes used for a map update). For a low network latency of 20ms, these judicious parameter settings in conjunction with our parallelized local map updating, reduce the 79% failure rate of the previous Edge SLAM systems down to 2%.

Network slicing is a key enabler of multi-tenancy in 5G-and-beyond networks that satisfies the distinct requirements of different use-cases. As the density of tenants increases over time, admission requests may be put in waiting queues leading to impatient tenant behaviors. Due to such behaviors, tenants may frequently leave-and-join the slice admission queues in search for an alternate mobile network provider (MNO). This can be a severe problem when slices are leased and released on a short-term basis. In this paper, we argue that the instant behavior of a slice may deviate considerably from the predicted average behavior known to the tenant through a slice controller and thus gives rise to impatient tenant behaviors. To address this problem, we propose Mutes, a multi-tenant switching algorithm, that aids tenants in finding the best MNO. For a fixed number of tenants, we show that Mutes attains a Nash Equilibrium. We also show that Mutes stabilizes the system under strict admission conditions in scenarios where tenants are allowed to randomly move between MNOs. Through simulations, we justify that the proposed Mutes algorithm significantly improves the resource assignment performance and converges faster than state-of-the-art policies.

Emerging applications of the Internet of Vehicles (IoV) require the wireless transmission of growing amounts of data, e.g., vehicle location and sensor data, over unreliable and increasingly congested wireless links between the mobile vehicles and the Road Side Units (RSUs); also, urban areas are becoming increasingly congested with vehicle road traffic. Road traffic management and data network traffic management to address these challenges require accurate representations of the road and network traffic, which are difficult due to the wide temporal and spatial correlations in the road and network traffic. We address this representation problem by designing, implementing, and evaluating the VeNet deep learning system to exploit the wirelessly transmitted data to predict future vehicle locations and network traffic. We develop the novel VeNet hybrid learning system that employs a stacked autoencoder (AE) consisting of a central AE and multiple local AEs that jointly feed into a Long-Short Term Memory (LSTM). We propose a new training algorithm for the hybrid VeNet learning system. The novel VeNet hybrid learning system conducts spatial learning that accounts for the spatial and temporal correlations in the dataset gathered from the mobile vehicles. Evaluations that involve measurements with custom-made Raspberry Pi vehicles indicate that the VeNet learning model significantly reduces the required signalling network traffic and prediction errors (down to approx. three quarters) compared to existing prediction models. At the same time, VeNet reduces the energy consumption on the vehicles as well as the learning delay.

Emerging microservices demand flexible low-latency processing of network functions in virtualized environments, e.g., as containerized network functions (CNFs). While ensuring highly responsive low-latency CNF processing, the computing environments should conserve energy to reduce costs. In this systems integration study, we develop and evaluate the novel XDP-Monitoring Energy-Adaptive Network Functions (X-MAN) framework for managing the CPU operational states (P-states) so as to reduce the power consumption while prioritizing low-latency service. Architecturally, X-MAN consists of lightweight traffic monitors that are attached to the virtual network interfaces in the kernel space for per-CNF traffic monitoring and a power manager in user space with a global view of the CNFs on a CPU core. Algorithmically, X-MAN monitors the CPU core utilization via hybrid simple and weighted moving average prediction fed by the traffic monitors and a power management based on step-based CPU core frequency (P-state) adjustments. We evaluate X-MAN through extensive measurements in a real physical testbed operating at up to 10 Gbps. We find that X-MAN incurs significantly shorter and more consistent monitoring latencies for the CPU utilization than a state-of-the-art CPU hardware counter approach. Also, X-MAN achieves more responsive CPU core frequency adjustments and more pronounced reductions of the CPU power consumption than a state-of-the-art code instrumentation approach. We make the X-MAN source code publicly available.

We propose a novel Federated Edge Network Utility Maximization (FEdg-NUM) architecture for solving a large-scale distributed network utility maximization (NUM) problem. In FEdg-NUM, clients with private utilities communicate to a peer-to-peer network of edge servers. This represents a departure from the classical distributed NUM master-slave configuration and enables distributed computing harnessing local communications. Compared to a solution using cloud synchronization via Ring AllReduce, we prove that our federated edge computing model has shorter run-time in the presence of network congestion, thanks to its configuration and its ability to make progress in the presence of intermittent links. The paper studies its convergence and run-time performance both analytically and numerically, and illustrates several possible networking applications.

This article presents a novel cloud-based charging management system for electric vehicles (EVs). Two levels of cloud computing, i.e., local and remote clouds, are employed to meet the different latency requirements of the heterogeneous EVs while exploiting the lower-cost computing in remote clouds. Specifically, we consider time-sensitive EVs at highway exit charging stations and EVs with relaxed timing constraints at parking lot charging stations. We propose algorithms for the interplay among EVs, charging stations, system operator, and clouds. Considering the contention-based random access for EVs to a 4G Long-Term Evolution network, and the quality of service metrics (average waiting time and blocking probability), the model is composed of: queuing-based cloud server planning, capacity planning in charging stations, delay analysis, and profit maximization. We propose and analyze aprice-incentive methodthat shifts heavy load from peak to off-peak hours, acapacity expansion methodthat accommodates the peak demand by purchasing additional electricity, and a hybrid method of price incentives and capacity expansion that balances the immediate charging needs of customers with the alleviation of the peak power grid load through price-incentive based demand control. Numerical results demonstrate the effectiveness of the proposed methods and elucidate the tradeoffs between the methods.

Managing cache content at the edge is one of the many use cases of 5G-and-beyond networks. However, increasing the density of Edge Data Centers (EDCs) to service requests is a crucial problem. To overcome this problem, recent research has advanced mobile device architecture paradigms and the content caching in a Mobile Device Cloud (MDC). These two service locations (EDCs and MDC) are registered with the Mobile Network Operator (MNO), enabling the MNO to control the content placement for profit maximization. As the user demands for content items are directly related to the QoS perceived by the user, it is important to understand the future popularity of the content items and to place them appropriately. Additionally, privacy issues have increased over time because of sensitive user information being divulged at the MDC. To preserve privacy, a branch of machine learning called Federated Learning (FL) can train machine learning models leaving the data in the end user devices. The paper contributions are as follows: (1) We introduce an FL algorithm called FedCo, that trains a deep-neural network (DNN) to predict the user demand of a specific content, so as to manage the content files placement at EDC and MDC sites. (2) We then conduct a theoretical evaluation of user demand behavior via prospect theory to justify revenue maximization for an MNO. (3) We show numerically via a multimedia content delivery use-case how the proposed model compares favorably with two state-of-the-art designs.

The increasing computing capabilities of client devices and the increasing demands for ultra-low latency services make it prudent to migrate some micro-service container computations from the cloud and multi-access edge computing (MEC) to the client devices. The migration of a container image requires compression and decompression, which are computationally demanding. We quantitatively examine the hardware acceleration of container image compression and decompression on a client device. Specifically, we compare the Intel®Quick Assist Technology (QAT) hardware acceleration with software compression/decompression. We find that QAT speeds up compression by a factor of over 7 compared to the single-core GZIP software, while QAT speeds up decompression by a factor of over 1.6 compared to the multi-core PIGZ software. QAT also reduces the CPU core utilization by over 15% for large container images. These QAT benefits come at the expense of Input/Output (IO) memory access bitrates of up to 900 Mbyte/s (while the software compression/decompression does not require IO memory access). The presented evaluation results provide reference benchmark performance characteristics of the achievable latencies for container image instantiation and migration with and without hardware acceleration of the compression and decompression of container images.

In recent years, the amounts of network traffic targeted towards cloud data centers have fluctuated based on user requests. This traffic is bursty and requires a high degree of attention. Due to the variable nature of this traffic, some requests need to be re-allocated on-the-fly. Such circumstances result in performance degradations due to resource management. As appropriate solutions can be proposed only based on understanding the workload and the environment, Reinforcement Learning (RL) is a strategy that is predominantly used. Further, it has been shown that the Poisson arrival rates do not capture real-world burstiness. Thus, we mainly have a two-fold problem to address: (i) the traffic requires a new modelling approach that can characterize the burstiness, and (ii) balancing the load that can maximize the reward to the provider in such circumstances. In this paper, we propose a novel, yet simple traffic modelling that enables burst detection based on an index policy. We show that the throughput constraints play a crucial role in scheduling and our proposed RL technique produces reliable results in such a scenario. Our RL algorithm decides what instance of the request traffic needs to be processed so that the cloud provider can maximize its profit and the decisions made in hindsight are non-biased. We compare the proposed policy with two state-of-the-art approaches and draw key inferences as to why an index policy performs better in scenarios that demand RL. We observe over five times shorter average wait times while bursty workload crosses a saturation limit of 150% compared to conventional policies.

We describe a distributed framework for resource sharing problems that arise in communications, micro-economics, and various networking applications. In particular, we consider a hierarchical multi-layer decomposition for network utility maximization (ML-NUM), where functionalities are assigned to different layers. The proposed methodology creates solutions with central management and distributed computations to the resource allocation problems. In non-stationary environments, the technique aims to respond quickly to the dynamics of the network by decreasing delay by partially shifting the communication and computational burden to the network edges. Our main contribution is a detailed analysis under the assumption that the network changes are on the same time-scale as the convergence time of the algorithms used for local computations. Moreover, assuming strong concavity and smoothness of the users' objective functions, and under some stability conditions for each layer, we present convergence rates and optimality bounds for the ML-NUM framework. In addition, the main benefits of the proposed method are demonstrated with numerical examples.

The sharing of skills over the Internet enables professionals to democratize their expertise and skills without exhausting their availability, e.g., through excessive traveling. To enable this Internet of Skills, we present a novel Digital Twin (DT) platform for the remote control of machines with human-in-the-loop. The DT of a remotely controlled machine acts effectively as an inter-layer between the operator and the controlled machine, e.g., robot arm. The DT can be optimized for a particular application to interact with the operator with an intuitive low-latency interface and, on other side, to control and monitor the quality of the remote task. Essentially, the human operator controls the DT, while the DT controls the remote robot. This paper introduces the DT framework for the remote control. The human-machine-human control loop is split into Virtual Reality (VR), remote control, and robot control loops. The proposed framework achieves low latency visual feedback and very short system reaction times for unexpected changes with arbitrary distances between operator and robot. Within the DT framework, this paper proposes a robot control algorithm for controlling time-critical robot applications over networks with considerable delays and jitter. The proposed framework has been implemented in a demonstrator with a robot arm and its DT in VR.

By moving computation and caching to the network edge, Mobile Edge Computing (MEC) offloads core networks and shortens data access latencies, which is important for large scale mobile multimedia services. Increasing the density of edge data centers to service these multimedia requests is uneconomical. Recent research has proven the benefits of involving devices in the delivery of multimedia services. This is done by exploiting the idle computation and storage resources via device-to-device (D2D) communication, i.e., by forming a so-called Mobile Device Cloud (MDC). Despite the flexibility and cost efficiency of this MDC paradigm, the timely allocation of caching resources to satisfy the dynamic user demands is challenging. This is mainly due to the uncertainty in resource availability of mobile devices. To this end, we propose Edge-Boost, a novel MDC caching architecture for lowlatency multimedia streaming services. We develop a novel fluid-based model to capture the dynamically changing network status. Additionally, we propose a dynamic caching allocation to jointly minimize caching cost and service latency. Edge-Boost achieves over 20% higher average cache utilization and 15% shorter average access latency than the state-of-the-art MDC approach.

Software-defined networking (SDN) and network function virtualization (NFV) processed in multi-access edge computing (MEC) cloud systems have been proposed as critical paradigms for achieving the low latency requirements of the tactile Internet. While virtual network functions (VNFs) allow greater flexibility compared to hardware-based solutions, the VNF abstraction also introduces additional packet processing delays. In this paper, we investigate the practical feasibility of NFV with respect to the tactile Internet latency requirements. We develop, implement, and evaluate Chain-based Low latency VNF ImplemeNtation (CALVIN), a low-latency management framework for distributed Service Function Chains (SFCs). CALVIN classifies VNFs into elementary, basic, and advanced VNFs; moreover, CALVIN implements elementary and basic VNFs in the kernel space, while the advanced VNFs are implemented in the user space. Throughout, CALVIN employs a distributed mapping with one VNF per Virtual Machine (VM) in a MEC system. Furthermore, CALVIN avoids the metadata structure processing and batch processing of packets in the conventional Linux networking stack so as to achieve short per-packet latencies. Our rigorous measurements on off-the-shelf conventional networking and computing hardware demonstrate that CALVIN achieves round-trip times from a MEC ingress point via two elementary forwarding VNFs (one in kernel space and one in user space) and a MEC server to a MEC egress point on the order of 0.32 ms. Our measurements also indicate that MEC network coding and encryption are feasible for small 256 byte packets with an MEC latency budget of 0.35 ms; whereas, large 1400 byte packets can complete the network coding, but not the encryption within the 0.35 ms.

Communication networks are the key enabling technology for our digital society. In order to sustain their critical services in the future, communication networks need to flexibly accommodate new requirements and changing contexts due to emerging diverse applications. In contrast to traditional networking technologies, software-oriented networking concepts, such as software-defined networking (SDN) and network function virtualization (NFV), provide ample opportunities for highly flexible network operations, enabling fast and simple adaptation of network resources and flows. This paper identifies the opportunities and challenges of adaptable softwarized networks and introduces a conceptual framework for adaptations in softwarized networks. We first explain how softwarized networks contribute to network adaptability through the functional primitives observation, composition, and control. We review the wide range of options for fine-granular observations as well as fine-granular composition and control provided by SDN and NFV. The multitude of fine-granular “tuning knobs” in adaptable softwarized networks complicates the decision making, which is the main focus of this paper. We propose to enhance the functional primitives observation, composition, and control with data-driven decision making, e.g., machine learning modules, resulting in deep observation, composition, and control. The data-driven decision making modules can learn and react to changes in the environment, e.g., new flow demands, so as to support meaningful decision making for adaptation in softwarized networks. Finally, we make the case for employing the concept of empowerment to realize truly “self-driving” networks.

Software Defined Networking (in short SDN, which is also an acronym for Software Defined Network), has emerged as a response to the limitations and complexities of traditional network architectures. At the heart of SDN lies the idea to consolidate the control over network devices into a logically centralized (software) controller separated from the data plane. The separation of the control plane and the data plane is realized via an open programming interface between the data plane switches and the SDN controller. The decoupling allows the control plane to evolve independently of the data plane, which enables faster innovation since software often exceeds hardware in innovation speed. Furthermore, logical centralization has the potential to simplify network operation and management by providing a single focal point where the consequences of management actions can be assessed, and possibly rejected if they would lead to some violation of operational constraints. OpenFlow, the standard SDN protocol today, is based on a simple match-action paradigm which results in great flexibilities, e.g., in terms of traffic engineering, definition of flows, as well as in-band network control functionalities.

Adaptive HTTP streaming requires a video to be encoded at multiple representations, that is, different qualities. Encoding these multiple representations is a computationally complex process, especially when using the recent High Efficiency Video Coding (HEVC) standard. In this paper, we consider a multi-rate HEVC encoder and identify four types of encoding information that can be reused from a high-quality reference encoding to speed up lower quality-dependent encodings. We show that the encoding decisions from the reference cannot be directly reused, as this would harm the overall rate-distortion (RD) performance. Thus, we propose methods to use the encoding information to constrain the RD optimization of the dependent encodings so that the encoding complexity is reduced while the RD performance is kept high. We additionally show that the proposed methods can be combined, leading to an efficient multi-rate encoder that exhibits high RD performance and substantial complexity reduction. Results show that the encoding time for 12 representations at different spatial resolutions and signal qualities can be reduced on average by 38%, while the average bitrate increases by less than 1%.

Rising energy costs, losses in the present-day electricity grid, risks from nuclear power generation, and global environmental changes are motivating a transformation of the conventional ways of generating electricity. Globally, there is a desire to rely more on renewable energy resources (RERs) for electricity generation. RERs reduce greenhouse gas emissions and may have economic benefits, e.g., through applying demand side management with dynamic pricing so as to shift loads from fossil fuel-based generators to RERs. The electricity grid is presently evolving toward an intelligent grid, the so-called smart grid (SG). One of the major goals of the future SG is to move toward 100% electricity generation from RERs, i.e., toward a 100% renewable grid. However, the disparate, intermittent, and typically widely geographically distributed nature of RERs complicates the integration of RERs into the SG. Moreover, individual RERs have generally lower capacity than conventional fossil fuel-based plants, and these RERs are based on a wide spectrum of different technologies. In this article, we give an overview of recent efforts that aim to integrate RERs into the SG. We outline the integration of RERs into the SG along with their supporting communication networks. We also discuss ongoing projects that seek to integrate RERs into the SG around the globe. Finally, we outline future research directions on integrating RERs into the SG.

Video cameras are increasingly used to provide real-time feedback in automatic control systems, such as autonomous driving and robotics systems. For such highly dynamic applications, the glass-to-glass (G2G) and glass-to-algorithm (G2A) latencies are critical. In this paper, we analyze the latencies in a point-to-point video transmission system and propose novel frame skipping and preemption approaches to reduce the G2G and G2A delays. We implement the proposed approaches in a prototype that shows significantly reduced G2G and G2A latencies as well as reduced transmission bitrate requirements compared with traditional video transmission schemes. In our low-delay video communication prototype, a VGA resolution video is transmitted with average G2G and G2A delays of 21.2 and 11.5 ms, respectively, with off-the-shelf hardware.

Random linear network coding (RLNC) has the potential to improve the performance of current and future Internet of Things (IoT) communication systems, but is computationally demanding due to matrix multiplications and inversions. Some single-core RLNC implementations achieve already sufficient coding speeds for contemporary multimedia streaming formats. However, advances in multimedia streaming formats and IoT applications will require the exploitation of heterogeneous multicore architectures, which are becoming common for a wide range of IoT nodes, including smartphones. In this paper, we introduce and evaluate efficient RLNC computing strategies for IoT node architectures, including the emerging heterogeneous big.LITTLE multicore architectures with multiple big (fast) cores and multiple LITTLE (slow) cores. In contrast to existing RLNC implementation strategies, we build on and adapt highly optimized dense matrix operations from the high performance computing field to RLNC on heterogeneous multicore IoT nodes. Our approach includes the optimization of RLNC matrix operations through optimized operations on matrix blocks with single instruction multiple data instructions. We schedule block operations on the heterogeneous cores through a directed acyclic graph that avoids artificial synchronization points while ensuring the data dependencies. We examine priority scheduling according to the number of outgoing dependencies of a task and data locality of cached blocks. Our extensive measurements with several heterogeneous big.LITTLE multicore IoT node and smartphone processor boards demonstrate higher RLNC encoding and decoding throughputs than existing approaches. Moreover, our measurements indicate that the utilization of more cores decreases energy consumption, which is an important goal for IoT nodes.

The papers in this special section focus on the deployment of information and communication technology (ICT) in smart grids as it relates to renewable energy resource management. The successful integration of renewable energy into the power grid is expected to reduce the dependence of the grid on the fossil fuels. The potential renewable energy resources include light, wind, vibration, heat, biofuel, biomass, and tides. It is envisaged that the use of renewable energy will reduce the use of traditional energy resources, such as nuclear, oil, and gas, in the future and this trend will continue in order to reduce the emission of greenhouse gases. The abundance of these renewable distributed energy resources (DERs) at the consumer side may help to develop distributed renewable energy generation at a large scale. The DERs will likely be an integral part of the future electric grid, i.e., the smart grid [4]–[6]. A prominent feature of the smart grid is that it allows for two-way communication between the utility and its customers through ICTs.

In a hybrid PON/xDSL access network, multiple customer premise equipment (CPE) nodes connect over individual digital subscriber lines (DSLs) to a drop-point device. The drop-point device, which is typically reverse powered from the customer, is co-located with an optical network unit (ONU) of the passive optical network (PON). We demonstrate that the drop-point experiences very high buffer occupancies when no flow control or standard Ethernet PAUSE frame flow control is employed. In order to reduce the buffer occupancies in the drop-point, we introduce two gated flow control protocols that extend the polling-based PON medium access control to the DSL segments between the CPEs and the ONUs. We analyze the timing of the gated flow control mechanisms to specify the latest possible time instant when CPEs can start the DSL upstream transmissions so that the ONU can forward the upstream transmissions at the full PON upstream transmission bit rate. Through extensive simulations for a wide range of bursty traffic models, we find that the gated flow control mechanisms, specifically, the ONU and CPE grant sizing policies, enable effective control of the maximum drop-point buffer occupancies.

Industrial networks demand centrally controlled quality of service (QoS), often in the form of hard real-time guarantees. Software-defined networking (SDN) provides a convenient paradigm for central QoS control. However, existing SDN-based solutions cannot guarantee hard real-time QoS as they rely on a control loop over the forwarding (data) and control planes. We propose a novel SDN-based QoS control framework that maintains an accurate network model through network calculus to avoid a control loop over forwarding and control planes, allocates resources to and routes flows over a network of “queue links,” whereby each physical network link houses multiple queue links (with different QoS levels), and manages QoS through a function split between delay-constrained least-cost routing on the network of queue links and the resource allocation to the queue links. This function split greatly reduces the computational complexity while achieving hard real-time QoS with high bandwidth utilization. Our evaluation results indicate that our function split approach allows for online runtime admission control and can achieve bandwidth utilization above 80% while meeting deterministic real-time QoS requirements.

Software defined networking (SDN) network hypervisors provide the functionalities needed for virtualizing software-defined networks. Hypervisors sit logically between the multiple virtual SDN networks (vSDNs), which reside on the underlying physical SDN network infrastructure, and the corresponding tenant (vSDN) controllers. Different SDN network hypervisor architectures have mainly been explored through proof-of-concept implementations. We fundamentally advance SDN network hypervisor research by conducting a model-based analysis of SDN hypervisor architectures. Specifically, we introduce mixed integer programming formulations for four different SDN network hypervisor architectures. Our model formulations can also optimize the placement of multi-controller switches in virtualized OpenFlow-enabled SDN networks. We employ our models to quantitatively examine the optimal placement of the hypervisor instances. We compare the control plane latencies of the different SDN hypervisor architectures and quantify the cost of virtualization, i.e., the latency overhead due to virtualizing SDN networks via hypervisors. For generalization, we quantify how the hypervisor architectures behave for different network topologies. Our model formulations and the insights drawn from our evaluations inform network operators about the trade-offs of the different hypervisor architectures and help choosing an architecture according to operator demands.

Small wireless cells have the potential to overcome bottlenecks in wireless access through the sharing of spectrum resources. However, current backhaul networks statically allocate resources, resulting in minuscule allocations when many small cells are connected to cellular operators with given resources. We introduce a novel access backhaul network architecture based on a smart gateway (Sm-GW) between the small cell base stations, e.g., long term evolution (LTE) enhanced Nodes B (eNBs), and the conventional backhaul gateways, e.g., LTE servicing/packet gateways (S/P-GWs). We specify the modest LTE protocol modifications that integrate the Sm-GW into the conventional LTE network. The Sm-GW flexibly schedules uplink transmissions for the eNBs. Our simulation evaluations indicate that the Sm-GW scheduling can fairly allocate uplink transmission bitrates to the eNBs and reduce packet delays. Based on software defined networking (SDN), we introduce a management mechanism that allows multiple operators, i.e., multiple S/P-GWs, to flexibly inter-operate via multiple Sm-GWs with a multitude of small cells. An SDN orchestrator coordinates the adaptive allocation of uplink transmission bitrates to Sm-GWs (which in turn allocate the uplink transmission bitrates to eNBs based on their demands). We formulate optimization problems for the operator (S/P-GW) resource allocation to Sm-GWs without and with sharing among operators. Our numerical evaluations indicate that the flexible SDN orchestration substantially increases the network throughput compared to the current static resource allocations.

Industrial networks demand centrally controlled Quality of Service (QoS), often in the form of hard real-time guarantees. We propose a novel SDN-based QoS control paradigm that (i) maintains an accurate network model through network calculus to avoid a control loop over forwarding and control planes, (ii) routes flows over a network of "queue links", whereby each physical network link houses multiple queue links (with different QoS levels), and (iii) manages QoS through delay-constrained least-cost (DCLC) routing on the network of queue links.

Fiber-Wireless (FiWi) networks have received considerable attention in the research community in the last few years as they offer an attractive way of integrating optical and wireless technology. As in every other type of networks, routing plays a major role in FiWi networks. Accordingly, a number of routing algorithms for FiWi networks have been proposed. Most of the routing algorithms attempt to find the “shortest path” from the source to the destination. A recent paper proposed a novel path length metric, where the contribution of a link towards path length computation depends not only on that link but also every other link that constitutes the path from the source to the destination. In this paper we address the problem of computing the shortest path using this path length metric. Moreover, we consider a variation of the metric and also provide an algorithm to compute the shortest path using this variation. As multipath routing provides a number of advantages over single path routing, we consider disjoint path routing with the new path length metric. We show that while the single path computation problem can be solved in polynomial time in both the cases, the disjoint path computation problem is NP-complete. We provide optimal solution for the NP-complete problem using integer linear programming and also provide two approximation algorithms with a performance bound of 4 and 2 respectively. The experimental evaluation of the approximation algorithms produced a near optimal solution in a fraction of a second.

The demand for bandwidth in broadband access networks has been increasing exponentially in recent years, due to high-bandwidth applications, such as video on demand (VoD), and peer-to peer (P2P) networking through increasingly powerful mobile devices. The demand for bandwidth and mobility has motivated the development of fiber-wireless (FiWi) networks, which combine the optical access network with the wireless network. FiWi networks support high bandwidth with optical networking mechanisms and mobility through wireless networking. In our previous work, we introduced clustered and localized routing (CluLoR) in FiWi networks. We examined CluLoR, which improves throughput-delay performance compared to a flat (unclustered) topology, only for an elementary dynamic bandwidth allocation (DBA) algorithm in the optical part of the FiWi network. In this paper, we extend our previous work by conducting an extensive evaluation of DBA algorithms on the CluLoR performance in FiWi networks. In this paper, our focus is how the wireless upstream traffic based on CluLoR going into the optical access network performs when conventional wired traffic is included in the optical network. We evaluate the delay performance for different practical scenarios and examine the underloading and overloading of the wireless and wired traffic respectively on FiWi networks. Our evaluations indicate that the performance impact of DBA algorithms depends on the source of the traffic and we provide comparisons in this paper.

Current Gigabit-class passive optical networks (PONs) evolve into next-generation PONs, whereby high-speed 10+ Gb/s time division multiplexing (TDM) and long-reach wavelength-broadcasting/routing wavelength division multiplexing (WDM) PONs are promising near-term candidates. On the other hand, next-generation wireless local area networks (WLANs) based on frame aggregation techniques will leverage physical-layer enhancements, giving rise to Gigabit-class very high throughput (VHT) WLANs. In this paper, we develop an analytical framework for evaluating the capacity and delay performance of a wide range of routing algorithms in converged fiber-wireless (FiWi) broadband access networks based on different next-generation PONs and a Gigabit-class multiradio multichannel WLAN-mesh front end. Our framework is very flexible and incorporates arbitrary frame size distributions, traffic matrices, optical/wireless propagation delays, data rates, and fiber faults. We verify the accuracy of our probabilistic analysis by means of simulation for the wireless and wireless-optical-wireless operation modes of various FiWi network architectures under peer-to-peer, upstream, uniform, and nonuniform traffic scenarios. The results indicate that our proposed optimized FiWi routing algorithm (OFRA) outperforms minimum (wireless) hop and delay routing in terms of throughput for balanced and unbalanced traffic loads, at the expense of a slightly increased mean delay at small to medium traffic loads.

Toward the vision of complete fixed-mobile convergence, a plethora of wireless, integrated optical-wireless, multipath, and energy-aware routing algorithms were proposed for legacy EPON/WLAN-mesh based bimodal fiber-wireless (FiWi) broadband access networks. In this paper, we present the first comprehensive analytical framework for providing deeper insights into the capacity and delay performance of routing algorithms in next-generation FiWi networks based on emerging powerful optical and wireless technologies such as long-reach 10+ Gb/s TDM/WDM PONs and Gigabit-class VHT WLANs.

The increased propagation delay of future long-reach passive optical networks (LR-PONs) may lead to a significantly increased idle time and delay if optical network units (ONUs) use conventional report-grant mechanisms. Sophisticated and efficient bandwidth allocation mechanisms are required to cope with the imposed propagation delay in LR-PONs. In this study, we evaluate three dynamic bandwidth allocation (DBA) frameworks in terms of frame (packet) delay; namely, we consider conventional (interleaved) polling for traditional PON and two recently introduced scheduling paradigms for next generation LR-PON, i.e., multi-thread polling (MT-P) and real-time polling (RT-P). We enhance MT-P and RT-P by applying the just-in-time framework. Next, we provide an analytical framework for evaluating the end-to-end frame delay in our enhanced MT-P and RT-P frameworks. We compare their performance with conventional polling and double-phase polling and investigate their shortcomings and advantages in an LR-PON setting. The simulation results closely match the analysis for this framework. Also, our results indicate that RT-P significantly reduces frame delay in LR-PONs compared to MT-P and conventional polling frameworks.

The goal of the study was to explore middle school students' preferences for an animated engineering tutor, and investigate their rationales for their choices. 77 middle school students participated in the study, and provided their preferences and rationales on various dimensions of an animated engineering tutor such as gender, age, personality, and clothing. Results showed that for teaching engineering in a computer-based instructional module, students preferred an animated engineering tutor that was similar to their age, matching their own gender, with a fun personality, and that speaks slowly.

We present different video traffic models for H.264 variable bit rate (VBR) videos. We propose our models on top of the recent unified traffic model developed by Dai et al. [1], which presents a frame-level hybrid framework for modeling MPEG-4 and H.264 multi-layer VBR video traffic. We exploit the hierarchical predication structure inherent in H.264 for intra-GoP (group of pictures) analysis. We model the children frames by considering various combinations of the correlation between the parent frames in the prediction structure. Our simulations show that modeling using the hierarchical prediction structure indeed improves capturing the statistical features of the videos and prediction of network performance, without an increase in the complexity as compared to the unified traffic model by Dai et al. [1], which was shown earlier to be better than previous traffic models.

Designing low latency polling schemes is one of the most important parts for passive optical networks (PONs), particularly for long-reach PONs (LR-PON) which suffer from long propagation delays. Sophisticated and efficient bandwidth allocation mechanisms are required to cope with the imposed transmission delay in LR-PONs. In this work, we evaluate three dynamic bandwidth allocation methods in terms of transmission delay. Namely, we consider conventional or interleaved polling for traditional PON and two recently introduced scheduling paradigms for next generation LR-PON, i.e., multi-thread polling (MT-P) and real-time polling (RT-P). We examine various flavors of each scheduling method and investigate their shortcomings and advantages in a LR-PON setting. Furthermore, we provide an analytical framework for obtaining packet delay in an enhanced version of RT-P method. The simulation results highly match the analysis for this framework. Also, our results indicate that RT-P method significantly reduces frame delay in LR-PONs compared to MT-P and conventional polling methods.

A gap between theory and practice in signals and systems courses is often reported at many universities as a key problem in recruiting signals and systems students. On the other hand, instructors often cite a lack of fundamental understanding in mathematics as an issue in this course. Students seem to be discontent with some of the abstraction of the signals and systems courses. In this work-in-progress paper, we describe a new pathways concept we introduced to address these problems by introducing in-depth discussions, several applications and hands-on exercises.

We introduce a convenient notational framework for Dynamic Bandwidth Allocation (DBA) algorithms in Ethernet Passive Optical Networks (EPONs) that uses the three principal axes of grant scheduling framework, grant sizing, and grant scheduling policy. We conduct comprehensive stability limit and packet delay investigations to determine which components have the strongest impact on these measures. We find that the grant sizing has the strongest impact on the delay and the combined grant scheduling framework and policy have the strongest impact on the stability limit. Further, we find that among the wide set of DBA algorithms we examined the shortest propagation delay first grant scheduling policy coupled with the limited with excess distribution grant sizing provides both the lowest delay and highest stability limit. The performance of shortest propagation delay first grant scheduling policy coupled with the limited with excess distribution grant sizing exceeds that of the online scheduling framework with limited grant sizing.

This paper puts forth a straw man proposal for future diverse internets. Guided by a principle that we call Align-and-Decouple (AD), our AD protocol architecture design calls for lightweight transport protocols that allow for decoupled operation of networks that are either owned by different organizations, or are networks of different types/protocols. TCP's fragmentation and congestion control mechanisms modify the original data stream so significantly that all new networks are essentially designed to just carry “TCP transformed” traffic. In contrast, with the AD principle, an AD-API is defined to consist of atomic transfer and pipe communication services, which can be supported by different providers with their own specific networking technologies, and yet allow for internetworking.

Building on the Ethernet Passive Optical Network (EPON) and Gigabit PON (GPON) standards, Next-Generation (NG) PONs (i) provide increased data rates, split ratios, wavelengths counts, and fiber lengths, as well as (ii) allow for all-optical integration of access and metro networks. In this paper we provide a comprehensive probabilistic analysis of the capacity (maximum mean packet throughput) and packet delay of subnetworks that can be used to form NG-PONs. Our analysis can cover a wide range of NG-PONs through taking the minimum capacity of the subnetworks forming the NG-PON and weighing the packet delays of the subnetworks. Our numerical and simulation results indicate that our analysis quite accurately characterizes the throughput-delay performance of EPON/GPON tree networks, including networks upgraded with higher data rates and wavelength counts. Our analysis also characterizes the trade-offs and bottlenecks when integrating EPON/GPON tree networks across a metro area with a ring, a Passive Star Coupler (PSC), or an Arrayed Waveguide Grating (AWG) for uniform and non-uniform traffic. To the best of our knowledge, the presented analysis is the first to consider multiple PONs interconnected via a metro network.

The scalable video coding extension (SVC) of the H.264/AVC standard is widely considered for IPTV. SVC supports a variety of scalability modes, including temporal, spatial as well as coarse-grain and medium-grain quality scalabilities. In this paper, we first give an overview of coarse-grain quality scalability (CGS). We generate traces of CGS encodings of long CIF resolution video sequences; the traces provide a simple yet effective characterization of CGS encoded video for performance evaluation of video transport systems, including IPTV systems. We conduct a detailed statistical analysis of the CGS video traces. We compare the bit rate-distortion (RD) and the bit rate variability-distortion (VD) performances of scalable CGS encodings with those of non-scalable SVC single layer encodings. We thus quantify the tradeoff between the rate adaptability afforded by CGS encoding and the cost in terms of RD efficiency compared to non-scalable single-layer video.

Due to the geographic distribution of its subscribers, Ethernet Passive Optical Networks (EPONs) have typically varying propagation delays between the Optical Network Units (ONUs) and the Optical Line Terminal (OLT). In this paper, we consider EPONs with an offline scheduling framework, which enables Quality-of-Service mechanisms by collecting bandwidth requests from all ONUs before the OLT makes dynamic bandwidth allocations for transmissions on the shared ONUs-to- OLT upstream channel. We propose and evaluate the Shortest Propagation Delay (SPD) first scheduling policy which sequences the ONUs' upstream transmissions in increasing order of the ONUs' propagation delays, i.e., the upstream transmission of the ONU with the smallest propagation delay is scheduled first. We formally analyze the competitiveness of SPD first scheduling and find that it achieves very close to optimal performance. We characterize the stability limit for Gated and Limited grant sizing in conjunction with SPD grant scheduling. We evaluate the cycle length and packet delay with SPD scheduling through probabilistic analysis and simulations and find significant reductions in packet delay with SPD first scheduling in EPONs with heterogeneous propagation delays, especially when Limited grant sizing is employed.

Packet-switched unidirectional and bidirectional ring wavelength division multiplexing (WDM) networks with destination stripping provide an increased capacity due to spatial wavelength reuse. Besides unicast traffic, future destination stripping ring WDM networks also need to support multicast traffic efficiently. This article examines the largest achievable transmitter throughput, receiver throughput, and multicast throughput of both unidirectional and bidirectional ring WDM networks with destination stripping. A probabilistic analysis evaluates both the nominal capacity, which is based on the mean hop distances traveled by the multicast packet copies, and the effective capacity, which is based on the ring segment with the highest utilization probability, for each of the three throughput metrics. The developed analytical methodology accommodates not only multicast traffic with arbitrary multicast fanout but also unicast and broadcast traffic. Numerical investigations compare the nominal transmission, receiver, and multicast capacities with the effective transmission, receiver, and multicast capacities and examine the impact of number of ring nodes and multicast fanout on the effective transmission, reception, and multicast capacity of both types of ring networks for different unicast, multicast, and broadcast traffic scenarios and different mixes of unicast and multicast traffic. The presented analytical methodology enables the evaluation and comparison of future multicast-capable medium access control (MAC) protocols for unidirectional and bidirectional ring WDM networks in terms of transmitter, receiver, and multicast throughput efficiency.

We propose an improved online scheduler for multichannel or Wavelength Division Multiplexed (WDM) Ethernet Passive Optical Network (EPON) upstream transmission. This scheduler employs a just-in-time online scheduling framework to increase the number of Optical Network Units (ONUs) that can be scheduled concurrently. We outline the overall structure of this scheduling framework and discuss adapting offline scheduling policies for use in this framework. We compare the average queueing delay performance of different schedulers that follow this new framework to a simple online scheduler that schedules ONUs as soon as their REPORTS are received at the Optical Line Terminal (OLT). Further, we show how this framework can be used to provide differentiated service to ONUs without waiting for all ONUs to REPORT. We conclude with some remarks regarding our performance findings and possibilities for future research.

We examine a recently proposed multichannel upgrade of optical single-channel ring networks where a subset of ring nodes is WDM upgraded and interconnected by a single-hop star WDM subnetwork in a pay-as-you-grow fashion. This evolutionary approach not only allows for fast and efficient multiple-failure recovery but also is well suited to efficiently sustain unpredictable changes and shifts in traffic loads. In this paper, we analytically investigate the maximum achievable capacity of the WDM star subnetwork upgrade of optical single-channel networks under a variety of unicast and multicast traffic scenarios and compare it to that of conventional WDM ring networks. In our analysis, we take priority of ring in-transit traffic, destination stripping, and maximum spatial reuse into account. Our findings show that under multicast traffic the configuration of the star subnetwork plays an important role in order to achieve high multicast capacity. Furthermore, under multicast traffic WDM upgrading and interconnecting a subset of ring nodes might be sufficient to achieve a larger multicast capacity than in WDM rings

In this paper we examine the relationship between the effective capacity (stability limit) of unidirectional and bidirectional packet-switched wavelength division multiplexing (WDM) ring networks for multicast traffic. We consider both bidirectional rings with one packet copy transmission per wavelength channel and two packet copy transmissions. We first prove bounds for the ratio of the multicast capacity of the bidirectional ring to the multicast capacity of the unidirectional ring. Specifically, we show that this ratio is at least two for two copy transmission in the bidirectional ring, and at most two for one copy transmission. We derive closed form expressions of the multicast capacity ratios for networks with a large number of nodes and from these expressions show that the ratios tend to two for a large number of multicast destinations. We demonstrate that for the bidirectional ring with two copy transmission the ratio becomes as large as 2.276. We also find that in the bidirectional ring, the capacity gain with two copy transmission over one copy transmission reaches 30.4%

Abstract — The development of pricing schemes that account for the specific challenges in streaming video to wireless clients is one of the key requirements for making wireless video services economically viable. In this paper we develop a conceptual framework for the pricing of wireless video streaming. Our framework incorporates the quality of the delivered video in the given networking context in an earnings model and captures the costs for the video service in a cost model. We discuss these models in the context of cellular, WLAN, and multi-hop wireless networks. We illustrate the developed pricing framework through numerical experiments with videos of a range of quality levels. Index Terms — wireless, video, multimedia, pricing I.

Currently available video traces for scalable encoded video with more than one layer are a convenient representation of the encoded video for the evaluation of networking mechanisms. The video distortion (RMSE) or quality (PSNR) for individual video frames in these traces, however, only allow for the calculation of the video quality of correctly received video frames; for lossy network transport, only a rough approximation can be made. With the availability of scalable offset distortion traces, which we introduce and evaluate in this paper, networking researchers are enabled to accurately calculate the video quality of scalable encoded video as it is perceived by the receiving client after lossy network transport.

We examine the bit rate variability-distortion (VD) curve of I, P, and B frames of MPEG-4 VBR encoded video sequences. We show that the concave VD curve shape at high compression ratios or large quantization scales, is influenced by both the texture and the motion information. We use linear and quadratic models for the texture and motion bits statistics and devise accurate VD curve models. The model parameters are obtained from statistics that are estimated from two encodings. This work extends our previous work on modeling the VD curve, which has applications for optimal statistical multiplexing of VBR streaming video

A variety of error resilience and scalable coding techniques have recently been proposed to facilitate the delivery of video over best-effort networks; a common drawback of these techniques is reduced compression efficiency. Also, MPEG-7 descriptors have recently been developed for the purpose of indexing. In this paper, we propose to employ MPEG-7 descriptors to improve the quality of the video delivered over best-effort networks. In particular, we propose a video transmission system that uses the motion activity descriptors to ensure robust video transmission. A novel motion activity extraction technique is proposed, which relies on a neural network approach. By considering several low-level visual features, our proposed extraction approach achieves high consistency with subjective evaluations of motion activities. In order to demonstrate the benefits of the proposed transmission system, we develop a selective packet dropping scheme that can be applied in case of network congestion. Simulations demonstrate that the reconstruction quality of the proposed congestion scheme can surpass conventional schemes by 1.2 dB. The network performance of the proposed transmission system when video sequences are coded into single layer or scalable layers is presented. We also present a transcoding scheme that achieves the optimal reconstructed quality by exploiting the motion activities of the underlying video sequence

Routing algorithms with time and message complexities that are provably low and independent of the total number of nodes in the network are essential for the design and operation of very large scale wireless mobile ad hoc networks (MANETs). In this paper, we develop and analyze Cluster Overlay Broadcast (COB), a low-complexity routing algorithm for MANETs. COB runs on top of a one-hop cluster cover of the network, which can be created and maintained using, for instance, the Least Cluster Change (LCC) algorithm. We formally prove that the LCC algorithm maintains a cluster cover with a constant density of cluster leaders with minimal update cost. COB discovers routes by flooding (broadcasting) route requests through the network of cluster leaders with a doubling radius technique. Building on the constant density property of the network of cluster leaders, we formally prove that, if there exists a route from a source to a destination node with a minimum hop count of A, then COB discovers a route with at most O(/spl Delta/) hops from the source to the destination node in at most O(/spl Delta/) time and by sending at Most O(/spl Delta//sup 2/) messages. We prove this result for arbitrary node distributions and mobility patterns and also show that COB adapts asymptotically optimally to the mobility of the nodes. In our simulation experiments, we examine the network layer performance of COB, compare it with Dynamic Source Routing, and investigate the impact of the MAC layer on COB routing.

Providing video services to heterogeneous clients in wireless ad hoc networks is particularly challenging as (i) the heterogeneous client processing and display capabilities typically prevent clients from processing and displaying the same encoded video information, and (ii) wireless connections typically suffer from bandwidth variability and transmission errors. We jointly address these two challenges by introducing a novel video coding strategy which combines multiple description coding, in particular temporal descriptors, with layered spatial coding. Our spatial scalable descriptor coding strategy enables heterogeneous clients in wireless multi-hop networks with path diversity to receive preencoded video streams over independent paths and to process only that amount of encoded video information that suits their processing and display capabilities. We evaluate our coding strategy through simulation experiments with a highly dynamic video sequence. We find that our coding strategy with two descriptors improves the quality of the received video by approximately 4 dB and cuts the quality variability approximately in half compared to layered coding.

Video traces containing the sizes and (PSNR) qualities of the individual frames of a video encoding are a convenient video representation for the evaluation of video networking mechanisms. These video traces can be used to find the frame loss probabilities of a lossy networking mechanism, but can not give the PSNR video quality after lossy network transport. To date the video quality after lossy network transport could only be determined through experiments with actual video or by approximating the quality of the frames affected by a loss with some low PSNR quality. In this paper we introduce and evaluate offset distortion traces with which the video quality after lossy network transport can be accurately determined without requiring experiments with actual video. We explain how the offset distortion traces can be used by networking researchers without equipment or experience in video signal processing to accurately evaluate video networking mechanisms in terms of the PSNR video quality.

Packet-switched unidirectional and bidirectional ring wavelength division multiplexing (WDM) networks with destination stripping provide an increased capacity due to spatial wavelength reuse. Besides unicast traffic, future destination stripping ring WDM networks also need to support multicast traffic efficiently. In this paper, we provide a probabilistic analysis of the mean hop distances traveled by multicast packet copies on the wavelength channels, and based on the mean hop distances analyze the nominal transmission capacity, reception capacity, and multicast capacity of both unidirectional and bidirectional ring WDM networks with destination stripping. The developed analytical methodology accommodates not only multicast traffic with arbitrary multicast fanout but also unicast and broadcast traffic. In our numerical investigations we examine the impact of number of ring nodes and multicast fanout on the transmission, reception, and multicast capacity of both types of ring networks for different unicast, multicast, and broadcast traffic scenarios and different mixes of unicast and multicast traffic. Our analytical methodology provides a foundation for extended analyses of the multicast capacity of WDM ring networks and enables the evaluation and comparison of future multicast-capable medium access control (MAC) protocols for unidirectional and bidirectional ring WDM networks in terms of transmitter, receiver, and multicast throughput efficiency.

Video services are expected to account for a large portion of the traffic in future wireless networks. Therefore, realistic traffic sources are needed to investigate the network performance of future communication protocols. Previously, we provided a publicly available library of frame size traces of long MPEG-4 and H.263 encoded videos in the QCIF format resulting in low bandwidth video streams. These traces can be used in 3G network simulations. Some future communication systems, such as WLAN systems, offer high data rates and therefore high quality video can be transmitted over such higher speed networks. We now present an addition to our existing trace library. For this addition we collected over 100 pre-encoded video sequences from the Web, generated the trace files, and conducted a thorough statistical evaluation. Because the pre-encoded video sequences are encoded by different users, their video settings differ in terms of codec, quality, format, and length. The advantage of user diversity for encoding is that it reflects very well the traffic situation in upcoming WLANs. Thus, the new traces are very suitable for the network performance evaluation of future WLANs.

Precision audio content description is one of the key components of next generation Internet multimedia search machines. We examine the usability of a combination of 39 different wavelets and three different types of neural nets for precision audio content description. More specifically, we develop a novel wavelet dispersion measure that measures obtained ranks of wavelet coefficients. Our dispersion measure in conjunction with a probabilistic radial basis neural network trained by only three independent example sets obtains a success rate of approximately 78% in identifying unknown complex classical music movements.

Wavelet-based encoding is now emerging as an efficient way to encode video for streaming over the Internet and for wireless applications. "Wavelet-based video coding has been recently added to the JPEG-2000 video standards. As wavelet encoded video emerges as the next generation video encoding method, it is vital to compare the efficiency of wavelet encoded video against the widely used DCT-based MPEG encoded video. However, due to the lack of long wavelet encoded video streams, most research has so far been based on short video traces. This paper presents a comparison study on MPEG vs wavelet encoded video traces for one hour movie excerpts with rate control. These long video sequences allow for the evaluation of long range dependency and self similarity of the generated video traffic, which has not been studied before in the context of comparing DCT and wavelet-based encoding. We focus on the elementary as well as self-similar traffic characteristics of the encoded video. A hump behavior for the variability of frame sizes is observed for increasing video bit rates for both wavelet and MPEG encoded video. In addition, the quality characteristics of the encoded video is examined and related to the traffic. Our results indicate that the wavelet encoded video results in higher video quality than MPEG encoded video. For the frame size variability we find different characteristics depending on the aggregation level for a given data rate. The results also indicate that the variation of quality resulting from the wavelet encoding is lower than for the MPEG encoded video

Single-hop WDM networks with a central passive star coupler (PSC), as well as single-hop networks with a central arrayed-waveguide grating (AWG) and a single transceiver at each node, have been extensively studied as solutions for the quickly increasing amounts of unicast and multicast traffic in the metropolitan area. The main bottlenecks of these networks are the lack of spatial wavelength reuse in the studied PSC based networks and the single transceiver in the studied AWG based metro WDM networks. In this paper we develop and evaluate the FTλ-FRλAWG network, which is based on a central AWG and has arrays of fixed-tuned transmitters and receivers at each node. Transceiver arrays are a mature technology, making the proposed network practical. In addition, the transmitter arrays allow for high speed signaling over the AWG while the receiver arrays relieve the receiver bottleneck arising from multicasting in conjunction with spatial wavelength reuse on the AWG. Our results from probabilistic analysis and simulation indicate that the FTλ-FRλAWG network gives particularly good throughput-delay performance for multicast traffic with small multicast group sizes or localized destination nodes, as well as for a mix of unicast and multicast traffic.

Both wavelength-division-multiplexing (WDM) networks with a ring architecture and WDM networks with a star architecture have been extensively studied as solutions to the ever increasing amount of traffic in the metropolitan area. Studies typically focus on either the ring or the star and significant advances have been made in the protocol design and performance optimization for the WDM ring and the WDM star, respectively. However, very little is known about the relative performance comparisons of ring and star networks. In this paper, we conduct a comprehensive comparison of a state-of-the-art WDM ring network with a state-of-the-art WDM star network. In particular, we compare time-slotted WDM ring networks (both single-fiber and dual-fiber) with tunable-transmitter and fixed-receiver (TT-FR) nodes and an arrayed-waveguide grating-based single-hop star network with tunable-transmitter and tunable-receiver (TT-TR) nodes. We evaluate mean aggregate throughput, relative packet loss, and mean delay by means of simulation for Bernoulli and self-similar traffic models for unicast traffic with uniform and hot-spot traffic matrices, as well as for multicast traffic. Our results quantify the fundamental performance characteristics of ring networks versus star networks and vice versa, as well as their respective performance limiting bottlenecks and, thus, provide guidance for directing future research efforts.

Single-hop WDM networks based on a central passive star coupler (PSC) or arrayed-waveguide grating (AWG) hub have received a great deal of attention as promising solutions for the quickly increasing traffic in metropolitan and local area networks. These single-hop networks suffer from a single point of failure: if the central hub fails, then all network connectivity is lost. To address this single point of failure in an efficient manner, we propose a novel single-hop WDM network, the AWG/spl par/PSC network. The AWG/spl par/PSC network consists of an AWG in parallel with a PSC. The AWG and PSC provide heterogeneous protection for each other; the AWG/spl par/PSC network remains functional when either the AWG or the PSC fails. If both AWG and PSC are functional, the AWG/spl par/PSC network uniquely combines the respective strengths of the two devices. By means of analysis and verifying simulations we find that the throughput of the AWG/spl par/PSC network is significantly larger than the total throughput obtained by combining the throughput of a stand-alone AWG network with the throughput of a stand-alone PSC network. We also find that the AWG/spl par/PSC network gives over a wide operating range a better throughput-delay performance than a network consisting of either two load sharing PSCs in parallel or two load sharing AWGs in parallel.

Research on multicasting in single-hop wavelength-division-multiplexing (WDM) networks has so far focused on networks based on the passive star coupler (PSC), a broadcast device. It has been shown that multicasting performance is improved by partitioning multicast transmissions into multiple multicast copies. However, the channel bottleneck of the PSC, which does not allow for spatial wavelength reuse, restricts the multicast performance. We investigate multicasting in a single-hop WDM network that is based on an arrayed-waveguide grating (AWG), a wavelength routing device that allows for spatial wavelength reuse. In our network, optical multicasting is enabled by wavelength-insensitive splitters that are attached to the AWG output ports. Multicasts are partitioned among the splitters and each multicast copy is routed to a different splitter by sending it on a different wavelength. We demonstrate that the spatial wavelength reuse in our network significantly improves the throughput-delay performance for multicast traffic. By means of analysis and simulations, we also demonstrate that, for a typical mix of unicast and multicast traffic, the throughput-delay performance is dramatically increased by transmitting multicast packets concurrently with control information in the reservation medium access control protocol of our AWG-based network.

Because Internet access rates are highly heterogeneous, many video content providers today make available different versions of the videos, with each version encoded at a different rate. Multiple video versions, however, require more server storage and may also dramatically impact cache performance in a traditional cache or in a CDN server. An alternative to versions is layered encoding, which can also provide multiple quality levels. Layered encoding requires less server storage capacity and may be more suitable for caching; but it typically increases transmission bandwidth due to encoding overhead. In this paper we compare video streaming of multiple versions with that of multiple layers in a caching environment. We examine caching and distribution strategies that use both versions and layers. Our analytical results indicate that mixed distribution/caching strategies provide the best overall performance.

Metropolitan WDM networks play an important role in the emerging Internet hierarchy; they interconnect the backbone WDM networks and the local access networks. The current SONET/SDH-over-WDM-ring metropolitan networks are expected to become a serious bottleneck - the so-called metropolitan gap-as they are faced with an increasing amount of bursty data traffic and quickly increasing bandwidths in the backbone networks and access networks. Innovative metropolitan WDM networks that are highly efficient and able to handle variable-size packets are needed to alleviate the metropolitan gap. In this paper we study an AWG-based single-hop WDM metropolitan network. We analyze the photonic switching of variable-size packets with spatial wavelength reuse. We derive computationally efficient and accurate expressions for the network throughput and delay. Our extensive numerical investigations - based on our analytical results and simulations - reveal that spatial wavelength reuse is crucial for efficient photonic packet switching. In typical scenarios, spatial wavelength reuse increases the throughput by 60% while reducing the delay by 40%.

The efficient distribution of stored information has become a major concern in the Internet which has increasingly become a vehicle for the transport of stored video. Because of the highly heterogeneous access to the Internet, researchers and engineers have argued for layered encoded video. We investigate delivering layered encoded video using caches. Based on the stochastic knapsack theory, we develop a model for the layered video caching problem. We propose heuristics to determine which videos and which layers in the videos should be cached in order to maximize the revenue from the streaming service. We evaluate the performance of our heuristics through extensive numerical experiments. We find that, for typical scenarios, the revenue increases nearly logarithmically with the cache size and linearly with the link bandwidth that connects the cache to the origin servers. We also consider service models with request queuing and negotiations about the delivered stream quality and find that both extensions provide only small revenue increases.

Continuous-media traffic (i.e., audio and video) can tolerate some loss but have rigid delay constraints. A natural QoS requirement for a continuous-media connection is a prescribed limit on the fraction of traffic that exceeds an end-to-end delay constraint. We propose and analyze a framework that provides such a statistical QoS guarantee to traffic in a packet-switched network. Providing statistical guarantees in a network is a notoriously difficult problem because traffic flows lose their original statistical characterizations at the outputs of queues. Our scheme uses bufferless statistical multiplexing combined with cascaded leaky buckets for smoothing and traffic contracting. This scheme along with a novel method for bounding the loss probability gives a tractable framework for providing end-to-end statistical QoS. Using MPEG video traces, we present numerical results that compare the connection-carrying capacity of our scheme with that of guaranteed service schemes (i.e., no loss) using GPS and RCS. Our numerical work indicates that our scheme can support significantly more connections without introducing significant traffic loss.

The efficient distribution of stored information has become a major concern in the Internet which has increasingly become a vehicle for the transport of stored video. Because of the highly heterogeneous access to the Internet, researchers and engineers have argued for layered encoded video. We investigate delivering layered encoded video using caches. Based on a stochastic knapsack model we develop a model for the layered video caching problem. We propose heuristics to determine which videos and which layers in the videos should be cached. We evaluate the performance of our heuristics through extensive numerical experiments. We also consider two intuitive extensions to the initial problem.

Streaming of continuous media over wireless links is a notoriously difficult problem. This is due to the stringent quality of service (QoS) requirements of continuous media and the unreliability of wireless links. We develop a streaming protocol for the real-time delivery of prerecorded continuous media from (to) a central base station to (from) multiple wireless clients within a wireless cell. Our protocol prefetches parts of the ongoing continuous media streams into prefetch buffers in the clients (base station). Our protocol prefetches according to a join-the-shortest-queue (JSQ) policy. By exploiting rate adaptation techniques of wireless data packet protocols, the JSQ policy dynamically allocates more transmission capacity to streams with small prefetched reserves. Our protocol uses channel probing to handle the location-dependent, time-varying, and bursty errors of wireless links. We evaluate our prefetching protocol through extensive simulations with VBR MPEG and H.263 encoded video traces. Our simulations indicate that for bursty VBR video with an average rate of 64 kb/s and typical wireless communication conditions our prefetching protocol achieves client starvation probabilities on the order of 10/sup -4/ and a bandwidth efficiency of 90% with prefetch buffers of 128 kbytes.

In order to provide quality of service (QoS) assurances networks perform call admission control before accepting a new connection. Rather than relying on a priori traffic descriptors (such as leaky buckets), which often poorly characterize the actual traffic, measurement-based admission control bases admission decisions on measurements of the actual traffic. We first develop a novel large deviations (LD) approach to measurement-based admission control for bufferless multiplexers. We then conduct simulation studies with traces of MPEG 1 encoded movies to compare the performance of the admission rules in the literature with that of the large deviations approach. We demonstrate that for bufferless multiplexing the LD approach achieves both higher link utilizations and smaller loss probabilities. Finally, we compare the performance of measurement-based admission control with that of traditional admission control, which relies on a priori traffic descriptors. Our numerical work indicates that measurement-based admission control achieves significant gains in link utilizations over traditional admission control.

Multimedia traffic can typically tolerate some loss but has rigid delay constraints. A natural QoS requirement for a multimedia connection is a prescribed bound on the the fraction of traffic that exceeds an end-to-end delay limit. We propose and analyze a traffic management scheme which guarantees QoS to multimedia traffic while simultaneously allowing for a large connection-carrying capacity. We study our traffic management scheme in the context of a single node. In order for the node to guarantee QoS, each connection's traffic is regulated. In order to support many connections, the link statistically multiplexes the connections' traffic. The scheme consists of (i) cascaded leaky-buckets for traffic regulation, (ii) smoothers at the ingresses, and (iii) bufferless statistical multiplexing within the node. For this scheme we show that loss probabilities are minimized with simple one-buffer smoothers which operate at specific minimum rates. We also show that the worst-case input traffic is extremal on-off traffic for all connections. These two results lead to a straightforward scheme for guaranteeing QoS to regulated traffic. Using MPEG video traces, we present numerical results which demonstrate the methodology. Finally, we compare the bufferless scheme with buffered statistical multiplexing.

We consider designing near video on demand (VoD) systems that minimize start-up latency while maintaining high image quality. Recently non-uniform segmentation has been used to develop periodic broadcasting techniques for near VoD. These techniques give significant reductions in start-up latency as compared with more conventional uniform segmentation. All of these schemes assume, however, that the videos are CBR-encoded. Since a CBR-encoded video has a larger average rate than an open-loop VBR encoding with the same image quality, there is potential to obtain further performance improvements by using VBR video. In this paper we develop a series of multiplexing schemes for the periodic broadcasting of VBR-encoded video, which are based on smoothing, server buffering and client prefetching. Two key but conflicting performance measures exist when using VBR video: latency and packet loss. By introducing small additional delays in our multiplexing schemes, our traced-based numerical work shows that the schemes can achieve nearly 100% link utilization with negligible packet loss. When the ratio of the CBR rate to the VBR average rate is a modest 1.8, start-up latency can be reduced by a factor of four or more for common scenarios.

We consider a finite-buffer packet multiplexer to which traffic arrives from several independent sources. The traffic from each of the sources is regulated, i.e., the amount of traffic that can enter the multiplexer is constrained by known regulator constraints. The regulator constraints depend on, the source and are more general than those resulting from cascaded leaky buckets. We assume that the traffic is adversarial to the extent permitted by the regulators. For lossless multiplexing, we show that if the original multiplexer is lossless it is possible to allocate bandwidth and buffer to the sources so that the resulting segregated systems are lossless. For lossy multiplexing, we use our results for lossless multiplexing to estimate the loss probability of the multiplexer. Our estimate involves transforming the original system into two independent resource systems, and using adversarial sources for the two independent resources to obtain a bound on the loss probabilities for the transformed system. We show that the adversarial sources are not extremal on-off sources, even when the regulator consists of a peak rate controller in series with a leaky bucket. We explicitly characterize the form of the adversarial source for the transformed problem. We also provide numerical results for the case of the simple regulator.

We present a high-performance prefetching protocol for the delivery of prerecorded VBR video from a server across a packet-switched network to a large number of clients. Not only does the protocol give constant perceptual quality and almost 100% link utilization, but it also allows for immediate commencement of the video upon user request and near instantaneous response to pause/resume and temporal jumps. The protocol requires: (1) that each client have a small amount of memory dedicated to the application; (2) that there is one bottleneck shared link between the server and the clients. Our protocol is based on the observation that there are frequent periods of time during which the shared link's bandwidth is under utilized. During these periods the server can prefetch frames from any of the ongoing videos and can send the frames to the buffers in the appropriate clients. The server chooses prefetched frames according to a join-the-shortest-queue policy. We present simulation results of our prefetch policy that are based on MPEG encoded traces.

We develop call admission policies for statistically multiplexing prerecorded sources over a bufferless transmission link. Our model is appropriate for video on demand, as well as other on-demand multimedia applications. In particular, we allow users to specify when the sources begin transmission; we also allow the user to invoke VCR actions such as pause and temporal jumps. We suppose that the quality of service (QoS) requirement allows for a small amount of packet loss. We develop a stochastic model which captures the random phases of the sources. We then apply large deviation theory to our model to develop global admission rules. The accuracy of the large deviation approximation is verified with simulation experiments employing importance sampling techniques. We also propose a refined admission rule which combines the global test and a myopic test. Numerical results are presented for the Star Wars trace; we find that the statistical multiplexing gain is potentially high and often insensitive to the QoS parameter. Finally, we develop efficient schemes for the real-time implementation of our global test. In particular, we demonstrate that the Taylor series expansion of the logarithmic moment generating function of the frame size distribution allows for fast and accurate admission decisions.