The research agenda on programmable data planes has been primarily focused on high-end networking devices, driven by technical requirements derived from operations & management needs of large scale datacenters and cloud providers. In this paper, we argue in favor of a yet incipient but equally paramount and challenging topic in this agenda: research on programmable low-end devices, like Low-power wide-area network (LPWAN). One main motivation is “unlocking” LPWAN, enabling one to freely redefine how they parse and process packets by means of Domain-specific languages such as P4. In addition to reducing capital expenditure by allowing interoperability between devices from multiple vendors, programmability would open LPWAN to an entire novel class of use cases, like providing inclusive internet access to technologically marginalized populations (such as rural communities). To contribute to this emerging research agenda, we propose a conceptual architecture and demonstrate the technical feasibility of a Programmable LPWAN by means of a proof-of-concept prototype, built using off-the-shelf hardware. More importantly, we present and discuss valuable lessons towards the design of such devices, maintaining their popular characteristics (like low power, low cost, long rage) yet freely (re)programmable for a broader class of novel use cases.

We present a prototype that could open the doors to mitigate digital exclusion in technologically underserved pop-ulations. By converging software-defined networks (SDN) with programmable data planes (PDP), our approach enables one to customize low-cost hardware to fit the network behavior. We integrate low-power wide-area networks (LPWANs), known for their scalability and efficiency, with PDPs to propose a flexible solution for changing requirements, such as distance and terrain configurations, validated through real-world test scenarios.

The rapid evolution of 5G and 6G technologies, coupled with growing environmental concerns, underscores the critical need for energy-efficient computer networks. To this end, a key challenge would be to minimize energy consumption by dynamically adjusting the number of active network devices based on the traffic demand and matrix. This requires an efficient mapping of the traffic matrix, which represents the amount of data traffic exchanged between different nodes over a given period, onto the network to ensure a minimal number of active while meeting performance requirements. Traditional approaches, based on Integer Linear Programming and heuristic algorithms, face significant limitations in scalability and computational efficiency, particularly for large-scale networks. To address these challenges, this work proposes a Machine Learning (ML)-based algorithm that leverages clustering techniques to identify near-optimal mappings of traffic matrices in Software-Defined Networks. Simulations on realistic network topologies demonstrate that our solution achieves substantial energy savings, up to 53 %, outperforms heuristic methods in execution time by orders of magnitude, and delivers near-optimal performance. These results highlight the potential of ML-driven approaches to enable scalable and energy-efficient network management.

As of today, TCP remains the de-facto transport protocol in the Internet. However, TCP may incur high delays, especially when retransmitting lost packets as they have to be retransmitted only by the source and after a timeout that is roughly equal to a round trip time. To reduce such delay, recent work [1]-[3] proposed to deploy a special network function, called Transport Assistant (TA), that is able to detect and retransmit lost TCP packets from inside the network rather than the source, and thereby, reduces retransmission delays. Unfortunately, there is no study so far on the impact of the placement of the TA on its performance benefits in terms of packet delivery delayIn this paper, we focus on the TA placement problem. We discuss the trade-offs and parameters to be considered to select the best placement for the TA. We mathematically model the TCP packet delivery delay, i.e., the time needed to deliver TCP packets when the TA is deployed, using different parameters like the location of the TA, the loss probabilities and the propagation delays of the network links. Thanks to this model, we study the impact of these parameters on the performance and efficiency of the TA.

Today’s Internet mainly relies on TCP protocol to ensure reliable communications between two endpoints. Unfortunately, this widely-used protocol may incur significant delay when transmitting lost packets. Indeed, with TCP, the source of the data detects lost packets using a timeout or duplicated acknowledgements before retransmitting them. As a result, the delay needed for a packet to reach the destination may become significant. This delay is estimated to be at least three times the end-to-end delay when the packet is lost once and could be even worse when the same packet is lost several times. As a matter of fact, this high delay cannot be tolerated by critical applications.To address this problem, in this paper, we focus on minimizing TCP retransmission delays and we introduce a novel network function called Transport Assistant that could be deployed within the network in order to cache, detect and retransmit lost packets. Thanks to this function, there is no need to wait for the source to detect and retransmit lost packets as the TA ensures packet retransmission from the network itself and thereby minimize retransmission delays. Through extensive experiments, we show that the TA allows to outperform the standard TCP by minimizing the average packet transmission time, the flow completion time, the packet loss and the number of retransmitted packets from the source.

Network softwarization is one of the key enablers of the future Internet evolution, also supporting the road from the fifth generation (5G) to the next-generation communication systems, namely 6G, with their main objective of bringing hyper-connected experience to every corner of society.

With the wide adoption of large-scale internet services and big data, the cloud has become the ideal environment to satisfy the ever-growing storage demand. In this context, data replication has been touted as the ultimate solution to improve data availability and reduce access time. However, replica management systems usually need to migrate and create a large number of data replicas over time between and within data centers, incurring a large overhead in terms of network load and availability. In this paper, we propose CRANE, an effiCient Replica migrAtion scheme for distributed cloud Storage systEms. CRANE complements any replica placement algorithm by efficiently managing replica creation in geo-distributed infrastructures in order to (1) minimize the time needed to copy the data to the new replica location, (2) avoid network congestion, and (3) ensure the minimum desired availability for the data. Through simulation and experimental results, we show that CRANE provides a sub-optimal solution for the replica migration problem with lower computational complexity than its integer linear program formulation. We also show that, compared to OpenStack Swift, CRANE is able to reduce by up to 60 percent the replica creation and migration time and by up to 50 percent the inter-data center network traffic while ensuring the minimum required data availability.

The Softwarization of networks is enabled by the SDN (Software-Defined Networking), NV (Network Virtualization), and NFV (Network Function Virtualization) paradigms, and offers many advantages for network operators, service providers and data-center providers. Given the strong interest in both industry and academia in the softwarization of telecommunication networks and cloud computing infrastructures, a series of special issues was established in IEEE Transactions on Network and Service Management, which aims at the timely publication of recent innovative research results on the management of softwarized networks.

Cloud storage services are gaining a widespread popularity thanks to their scalability and performance. Several companies and users are relying on such services to store and retrieve their files. In this context, the file transfer time is critical for users' satisfaction. This time can be minimized by carefully selecting the routing strategy to transfer the flow of packets associated with each file. In this paper, we introduce a novel class-based routing strategy called LUNA that is able to minimize the flow completion time. LUNA classifies the flows into mice and elephants based on their size. Afterwards, it leverages a machine learning technique called association rules to generate the forwarding rules and route each flow based on its class (i.e., mouse or elephant). Experimental results show that LUNA has successfully identified the class of 80% of the flows. Furthermore, its class-based routing outperforms basic routing strategies in terms of flow completion time, throughput and packet loss by almost 47%, 41% and 23%, respectively.

Network Function Virtualization is known for its ability to reduce deployment costs and improve the flexibility and scalability of network functions. Due to processing capacity limitation, the infrastructure provider needs to instantiate one or more instances of a particular network function when the amount of traffic increases. Most of network functions are stateful, which means that they keep a state that may be frequently read or updated (e.g., statistics like number of packets or bytes per flow). As a result, the instances of the same virtual network function should constantly share the same state to prevent incorrect operation. In this context, a major challenge is how to efficiently ensure the consistency among instances while minimizing communication cost for synchronizing their state and ensuring the synchronization delay does not exceed a certain bound set by the operator.In this paper, we propose a technique to identify the optimal communication pattern between the instances of the same network function in order to minimize their synchronization cost. Moreover, we propose to use a special network function named Synchronization Function to ensure consistency among a set of instances and to minimize the synchronization cost. We first mathematically model the problem of finding the optimal synchronization pattern and the optimal placement and number of synchronization functions as an integer linear program that minimizes the synchronization cost and ensures a bounded synchronization delay. Last, we put forward three algorithms to cope with large-scale scenarios of the problem. Extensive simulations show that the proposed algorithms efficiently find near-optimal solutions with minimal computation time.

Service Function Chain (SFC) provides an end-to-end service by processing traffic flow through a series of Virtual Network Functions (VNFs) in a specific order. Satisfying user's demands (e.g., end-to-end delay) on one hand and minimizing the cost of SFC deployment in terms of energy and resource on the other hand, introduces VNFs placement as a crucial issue that is receiving significant attention by researchers. To address this problem and boost the performance of SFC, different techniques such as Network Function (NF) distribution, NF parallelism and optimal resource allocation have been utilized. Applying these mechanisms imposes other costs which must be taken into account by network providers. In this paper, we introduce CoDeC as a Cost-effective and Delay-aware resource allocation approach. By having user defined end-to-end threshold and using aforementioned mechanisms, CoDeC tries to place the requested VNFs with the minimum cost of deployment, distribution, parallelism and energy. Therefore, we formulate the addressed problem in form of Mixed Integer linear Programming (MILP) model. We then show that the problem is NP-complete and suffers from high time complexity in large-scale scenarios. Thus, a heuristic algorithm is introduced to determine a near-optimal solution in a reasonable amount of time. Our simulation results show that CoDeC achieves better performance in term of cost and acceptance rate compared to using each mechanism individually.

With the emergence of software defined networking and network function virtualization technologies, network services are expected to be offered as service function chains made out from virtual network functions that are connected to steer and process the incoming traffic. In this context, achieving the survivability of these chains against failures is a key challenge to ensure high availability and continuity of the services. A promising solution proposed in the literature is to provision backups for the virtual network functions that could be shared among multiple service chains. These backups are used in case of a failure to take over the failed functions and ensure service continuity. In this paper, we propose two solutions to efficiently place and provision the shared backups in order to ensure the survivability of the service chains against single node failures. The originality of these solutions is that they leverage the migration of virtual network functions to minimize the resources consumed by the backups. Simulation results show that, compared to existing solutions, the proposed schemes leveraging migration are able to reduce by up to 20% the amount of resources allocated for the shared backups while ensuring the survivability of the service chains.

The softwarization of networks is enabled by the SDN (Software-Defined Networking), NV (Network Virtualization), and NFV (Network Function Virtualization) paradigms, and offers many advantages for network operators, service providers and datacenter providers. Given the strong interest in both industry and academia in the softwarization of telecommunication networks and cloud computing infrastructures, a series of special issues was established in IEEE Transactions on Network and Service Management, which aims at the timely publication of recent innovative research results on management of softwarized networks.

Network Function Virtualization (NFV) is an emergent paradigm that is currently transforming the way network services are provisioned and managed. The main idea of NFV is to decouple network functions from the hardware running them. This allows to reduce deployment costs and further improve the flexibility and the scalability of network services. Despite these benefits, a major challenge cloud providers are still facing is how to efficiently allocate resources for NFV-based services in a way that reduces operational costs and maximizes their profits. In this paper, we address this particular challenge and propose an effective profit-driven service chain provisioning scheme designed for large-scale infrastructures spanning different geographically-distributed sites. We hence propose three algorithms that maximize the provider's profit taking into consideration energy consumption of the infrastructure and the variability of energy prices in different locations. Through extensive simulations, we show that these algorithms are able to efficiently find near-optimal resource allocations and maximize the provider's profit with minimal computational complexity.

In the era of Big data, cloud storage services have become the option of choice to store and share data thanks to their cost-effectiveness and seemingly limitless capacity. The increasing success of these services is driving cloud providers to further improve their storage management systems in order to offer more stringent guarantees on data availability and access time. However, despite recent efforts towards this goal, existing solutions have largely overlooked the heterogeneity of the workloads and the underlying storage components in terms of failure rates, capacity and I/O speed. To fill this gap, we present in this paper a heterogeneity-aware data management scheme (dubbed Heron) based on a genetic algorithm that takes into consideration disk heterogeneity to satisfy SLA requirements in terms of access time and availability and minimizes costs in terms of data migration, storage and energy consumption. Through realistic simulations, we show that Heron significantly improves data availability and access time and ensures minimal storage costs and data migration overhead compared to heterogeneity-oblivious solutions.

MapReduce has become a popular model for large-scale data processing in recent years. However, existing MapRe-duce schedulers still suffer from an issue known as partitioning skew, where the output of map tasks is unevenly distributed among reduce tasks. In this paper, we present DREAMS, a framework that provides run-time partitioning skew mitigation. Unlike previous approaches that try to balance the workload of reducers by repartitioning the intermediate data assigned to each reduce task, in DREAMS we cope with partitioning skew by adjusting task run-time resource allocation. We show that our approach allows DREAMS to eliminate the overhead of data repartitioning. Through experiments using both real and synthetic workloads running on a 11-node virtual virtualised Hadoop cluster, we show that DREAMS can effectively mitigate negative impact of partitioning skew, thereby improving job performance by up to 20.3%.

MapReduce has become a popular model for data-intensive computation in recent years. By breaking down each job into small map and reduce tasks and executing them in parallel across a large number of machines, MapReduce can significantly reduce the running time of data-intensive jobs. However, despite recent efforts toward designing resource-efficient MapReduce schedulers, existing solutions that focus on scheduling at the task-level still offer sub-optimal job performance. This is because tasks can have highly varying resource requirements during their lifetime, which makes it difficult for task-level schedulers to effectively utilize available resources to reduce job execution time. To address this limitation, we introduce PRISM, a fine-grained resource-aware MapReduce scheduler that divides tasks into phases, where each phase has a constant resource usage profile, and performs scheduling at the phase level. We first demonstrate the importance of phase-level scheduling by showing the resource usage variability within the lifetime of a task using a wide-range of MapReduce jobs. We then present a phase-level scheduling algorithm that improves execution parallelism and resource utilization without introducing stragglers. In a 10-node Hadoop cluster running standard benchmarks, PRISM offers high resource utilization and provides 1.3× improvement in job running time compared to the current Hadoop schedulers.

With the massive adoption of cloud-based services, high energy consumption and carbon footprint of cloud infrastructures have become a major concern in the IT industry. Consequently, many governments and IT advisory organizations have urged IT stakeholders (i.e., cloud provider and cloud customers) to embrace green IT and regularly monitor and report their carbon emissions and put in place efficient strategies and techniques to control the environmental impact of their infrastructures and/or applications. Motivated by this growing trend, we investigate, in this paper, how cloud providers can meet Service Level Agreements (SLAs) with green requirements. In such SLAs, a cloud customer requires from cloud providers that carbon emissions generated by the leased resources should not exceed a fixed bound. We hence propose a resource management framework allowing cloud providers to provision resources in the form of Virtual Data Centers (VDCs) (i.e., a set of virtual machines and virtual links with guaranteed bandwidth) across a geo-distributed infrastructure with the aim of reducing operational costs and green SLA violation penalties. Extensive simulations show that the proposed solution maximizes the cloud provider's profit and minimizes the violation of green SLAs.

Cloud computing has become a cost-effective model for deploying online services in recent years. To improve the Quality-of-Service (QoS) of the provisioned services, recently a number of proposals have advocated to provision both guaranteed server and network resources in the form of Virtual Data Centers (VDCs). However, existing VDC scheduling algorithms have not fully considered the reliability aspect of the allocations in terms of (1) hardware failure characteristics on which the service is hosted, and (2) the impact of individual failures on service availability, given the dependencies among the virtual components. To address this limitation, in this paper we present a technique for computing VDC availability that considers heterogeneous hardware failure rates and dependencies among virtual components. We then propose Venice, an availability-aware VDC embedding framework for achieving high VDC availability and low operational costs. Experiments show Venice can significantly improve VDC availability while achieving higher income compared to availability-oblivious solutions.

With the proliferation of cloud computing, virtu-alization has become the cornerstone of modern data centers and an effective solution to reduce operational costs, maximize utilization and improve performance and reliability. One of the powerful features provided by virtualization is Virtual Machine (VM) migration, which facilitates moving workloads within the infrastructure to reach various performance objectives. As recent virtual resource management schemes are more reliant on this feature, a large number of VM migrations may be triggered simultaneously to optimize resource allocations. In this context, a challenging problem is to find an efficient migration plan, i.e., an optimal sequence in which migrations should be triggered in order to minimize the total migration time and impact on services. In this paper, we propose CQNCR (read as sequencer), an effective technique for determining the execution order of massive VM migrations within data centers. Specifically, given an initial and a target resource configuration, CQNCR sequences VM migrations so as to efficiently reach the final configuration with minimal time and impact on performance. Experiments show that CQNCR can significantly reduce total migration time by up to 35% and service downtime by up to 60%.

Recent advances in virtualization technology have made it possible to partition a network into multiple virtual networks managed by different users. Although virtual networks share the same physical infrastructure, they host diverse applications with different goals. Unfortunately, current virtual network provisioning solutions have only focused on achieving a single objective that may not be suited for all the applications deployed across the network. In this paper, we propose an adaptive provisioning framework for virtualized networks that takes into consideration the characteristics of multiple applications and their distinct performance objectives. The proposed framework is based on the concept of allocation paradigm, which is defined as a set of application-driven provisioning policies that guide the resource allocation process. To determine the efficiency of a particular paradigm, we propose a virtual network performance computation model based on data measured from existing benchmarks. Simulation results show that our model helps network providers to select the best allocation paradigms in terms of provisioning quality.

With the growing adoption of Software Defined Networking (SDN), there is a compelling need for SDN emulators that facilitate experimenting with new SDN-based technologies. Unfortunately, Mininet [1], the de facto standard emulator for software defined networks, fails to scale with network size and traffic volume. The aim of this paper is to fill the void in this space by presenting a low cost and scalable network emulator called Distributed OpenFlow Testbed (DOT). It can emulate large SDN deployments by distributing the workload over a cluster of compute nodes. Through extensive experiments, we show that DOT can overcome the limitations of Mininet and emulate larger networks. We also demonstrate the effectiveness of DOT on four Rocketfuel topologies. DOT is available for public use and community-driven development at dothub.org.

With the growing adoption of Software Defined Networking (SDN) technology, there is a compelling need for an SDN emulator that can facilitate experimenting with new SDN solutions. Unfortunately, Mininet, the de facto standard emulator for software defined networks, fails to scale with network size and traffic volume. To address these limitations, we developed Distributed OpenFlow Testbed (DOT), a highly scalable emulator for SDN. It can emulate large SDN deployments by distributing the workload over a cluster of compute nodes. Moreover, DOT can emulate a wider range of network services compared to other publicly available SDN emulators and simulators. Our demonstration will illustrate several features of DOT including: (i) how easy it is to setup the emulator, (ii) how to deploy a topology using a single configuration file, (iii) how to run a connectivity test to ensure that the emulated network is properly deployed, and (iv) how to control and monitor the emulated components from a centralized location. We will also showcase DOT by emulating two applications: (i) policy based traffic steering through middleboxes and (ii) traffic monitoring.

Data centers consume tremendous amounts of energy in terms of power distribution and cooling. Dynamic capacity provisioning is a promising approach for reducing energy consumption by dynamically adjusting the number of active machines to match resource demands. However, despite extensive studies of the problem, existing solutions have not fully considered the heterogeneity of both workload and machine hardware found in production environments. In particular, production data centers often comprise heterogeneous machines with different capacities and energy consumption characteristics. Meanwhile, the production cloud workloads typically consist of diverse applications with different priorities, performance and resource requirements. Failure to consider the heterogeneity of both machines and workloads will lead to both sub-optimal energy-savings and long scheduling delays, due to incompatibility between workload requirements and the resources offered by the provisioned machines. To address this limitation, we present Harmony, a Heterogeneity-Aware dynamic capacity provisioning scheme for cloud data centers. Specifically, we first use the K-means clustering algorithm to divide workload into distinct task classes with similar characteristics in terms of resource and performance requirements. Then we present a technique that dynamically adjusting the number of machines to minimize total energy consumption and scheduling delay. Simulations using traces from a Google's compute cluster demonstrate Harmony can reduce energy by 28 percent compared to heterogeneity-oblivious solutions.

Software Defined Networking (SDN) has emerged as a new paradigm that offers the programmability required to dynamically configure and control a network. A traditional SDN implementation relies on a logically centralized controller that runs the control plane. However, in a large-scale WAN deployment, this rudimentary centralized approach has several limitations related to performance and scalability. To address these issues, recent proposals have advocated deploying multiple controllers that work cooperatively to control a network. Nonetheless, this approach drags in an interesting problem, which we call the Dynamic Controller Provisioning Problem (DCPP). DCPP dynamically adapts the number of controllers and their locations with changing network conditions, in order to minimize flow setup time and communication overhead. In this paper, we propose a framework for deploying multiple controllers within an WAN. Our framework dynamically adjusts the number of active controllers and delegates each controller with a subset of Openflow switches according to network dynamics while ensuring minimal flow setup time and communication overhead. To this end, we formulate the optimal controller provisioning problem as an Integer Linear Program (ILP) and propose two heuristics to solve it. Simulation results show that our solution minimizes flow setup time while incurring very low communication overhead.

Data centers today consume tremendous amount of energy in terms of power distribution and cooling. Dynamic capacity provisioning is a promising approach for reducing energy consumption by dynamically adjusting the number of active machines to match resource demands. However, despite extensive studies of the problem, existing solutions for dynamic capacity provisioning have not fully considered the heterogeneity of both workload and machine hardware found in production environments. In particular, production data centers often comprise several generations of machines with different capacities, capabilities and energy consumption characteristics. Meanwhile, the workloads running in these data centers typically consist of a wide variety of applications with different priorities, performance objectives and resource requirements. Failure to consider heterogenous characteristics will lead to both sub-optimal energy-savings and long scheduling delays, due to incompatibility between workload requirements and the resources offered by the provisioned machines. To address this limitation, in this paper we present HARMONY, a Heterogeneity-Aware Resource Management System for dynamic capacity provisioning in cloud computing environments. Specifically, we first use the K-means clustering algorithm to divide the workload into distinct task classes with similar characteristics in terms of resource and performance requirements. Then we present a novel technique for dynamically adjusting the number of machines of each type to minimize total energy consumption and performance penalty in terms of scheduling delay. Through simulations using real traces from Google's compute clusters, we found that our approach can improve data center energy efficiency by up to 28% compared to heterogeneity-oblivious solutions.

Large-scale online service providers have been increasingly relying on geographically distributed cloud infrastructures for service hosting and delivery. In this context, a key challenge faced by service providers is to determine the locations where service applications should be placed such that the hosting cost is minimized while key performance requirements (e.g., response time) are ensured. Furthermore, the dynamic nature of both demand pattern and infrastructure cost favors a dynamic solution to this problem. Currently most of the existing solutions for service placement have either ignored dynamics, or provided solutions inadequate to achieve this objective. In this paper, we present a framework for dynamic service placement problems based on control- and game-theoretic models. In particular, we present a solution that optimizes the hosting cost dynamically over time according to both demand and resource price fluctuations. We further consider the case where multiple service providers compete for resources in a dynamic manner. This paper extends our previous work [1] by analyzing the outcome of the competition in terms of both price of stability and price of anarchy. Our analysis suggests that in an uncoordinated scenario where service providers behave in a selfish manner, the resulting Nash equilibrium can be arbitrarily worse than the optimal centralized solution in terms of social welfare. Based on this observation, we present a coordination mechanism that can be employed by the infrastructure provider to maximize the social welfare of the system. Finally, we demonstrate the effectiveness of our solutions using realistic simulations.

Cloud computing promises to provide on-demand computing, storage, and networking resources. However, most cloud providers simply offer virtual machines (VMs) without bandwidth and delay guarantees, which may hurt the performance of the deployed services. Recently, some proposals suggested remediating such limitation by offering virtual data centers (VDCs) instead of VMs only. However, they have only considered the case where VDCs are embedded within a single data center. In practice, infrastructure providers should have the ability to provision requested VDCs across their distributed infrastructure to achieve multiple goals including revenue maximization, operational costs reduction, energy efficiency, and green IT, or to simply satisfy geographic location constraints of the VDCs. In this paper, we propose Greenhead, a holistic resource management framework for embedding VDCs across geographically distributed data centers connected through a backbone network. The goal of Greenhead is to maximize the cloud provider's revenue while ensuring that the infrastructure is as environment-friendly as possible. To evaluate the effectiveness of our proposal, we conducted extensive simulations of four data centers connected through the NSFNet topology. Results show that Greenhead improves requests' acceptance ratio and revenue by up to 40 percent while ensuring high usage of renewable energy and minimal carbon footprint.

Intrusion Detection and/or Prevention Systems (IDPSs) are now a crucial defensive measure to defend against attacks intended to breach the security and operation of enterprise information systems. The IDPS configuration can, however, have a negative impact on network performance in terms of end-to-end delay and packet loss. This paper proposes an analytical queuing model based on the embedded Markov chain which analyzes the performance of the IDPS and evaluates its impact on performance. Through extensive simulations, we validate the proposed model and the numerical equations that estimate various performance metrics. Our results show that this model can be leveraged to assess and set up an effective configuration for the IDPS, achieving simultaneously the trade-off between security enforcement levels on one side and network Quality of Service (QoS) requirements on the other.

Large-scale online service providers have been increasingly relying on geographically distributed cloud infrastructures for service hosting and delivery. In this context, a key challenge faced by service providers is to determine the locations where service applications should be placed such that the hosting cost is minimized while key performance requirements (e.g. response time) are assured. Furthermore, the dynamic nature of both demand pattern and infrastructure cost favors a dynamic solution to this problem. Currently most of the existing solutions for service placement have either ignored dynamics, or provided inadequate solutions that achieve both objectives at the same time. In this paper, we present a framework for dynamic service placement problems based on control- and game-theoretic models. In particular, we present a solution that optimizes the desired objective dynamically over time according to both demand and resource price fluctuations. We further consider the case where multiple service providers compete for resource in a dynamic manner, and show that there is a Nash equilibrium solution which is socially optimal. Using simulations based on realistic topologies, demand and resource prices, we demonstrate the effectiveness of our solution in realistic settings.

The emergence of new kinds of applications and technologies (e.g., data-intensive applications, server virtualization) has led to a better utilization of the network resources. However, it has also led to more bandwidth consumption and more congestion especially inside data center networks. Thus, researchers are focusing again on TCP and Active Queue Management (AQM) mechanisms in order to better control congestion and to cope with application requirements in terms of end-to-end delay [1], [2], [3]. Recently, we proposed a new AQM mechanism (called α_SNFAQM) that uses traffic prediction to accurately detect future congestion and to proactively act upon it [4]. In this paper, we develop an analytical model to assess the effect of α_SNFAQM on TCP. The study proves that this AQM is efficient enough to stabilize queue size in routers/switches, and thereby allowing to control end-to-end packet delay. These results have been also validated by simulations for a topology with multiple bottleneck links. They show that α_SNFAQM outperforms other AQM schemes like RED, PAQM and APACE in stabilizing instantaneous queue length, while keeping a high utilization of the links and the same packet loss rate.

Optical Burst Switching (OBS) has been developed as an efficient switching technique for the next generation optical Internet. A critical issue for OBS networks is the burst loss which could occur due to contention and/or insufficient offset time. Burst Loss Ratio (BLR) is used as the main performance parameter in bufferless OBS networks. This paper proposes a new TCP statistics based method to predict the BLR without using any feedback information from the network. The idea is to estimate the BLR based on the TCP statistics available at the edge node. Our proposed BLR prediction method is then integrated into the closed loop feedback control model to control the BLR inside the network. Our simulation results clearly show that our proposed method improves the efficiency of the closed loop feedback control model while avoiding the use of any feedback information from the network.

Optical Burst Switching (OBS) has been developed as an efficient switching technique for the next generation optical Internet. A critical issue in OBS is the burst loss which could occur due to contention and/or insufficient offset time. Burst Loss Ratio (BLR) is used as the main performance parameter in OBS networks. In this paper, we investigate the assignment of the offset time and its effect on the measured end to end (E2E) delay. A novel feedback control technique is proposed to adapt the offset time based on the network condition in terms of BLR. Simulations show that the feedback control is able to adjust the offset automatically and dynamically. Hence, it reduces both the BLR due to insufficient offset and the E2E delay.

This paper proposes a novel scheme that uses feedback control approaches to support quality of service (QoS) in optical burst switching (OBS) networks. This work provides service level objectives in terms of burst loss ratio (BLR) for each class of bursts. The BLR is the ratio between the lost bursts to the sent bursts. Using feedback control approaches computes accurate burstification rate for each class of bursts. Burstification rates are computed at each burst manager controller for each class based on the previous measured value of the burst loss rate and the desired burst loss rate. Simulation results on NSFNET network have showed that the feedback control scheme guarantees a BLR to hover around the desired value for each class of bursts.

This paper considers the use of closed loop feedback control theoretic techniques to improve the performance of optical burst switching (OBS) networks. In OBS networks, the burst loss ratio (BLR) is the ratio between the lost bursts to the sent bursts. The BLR is used as a performance metric. The desired burst loss ratio depends on the application using the network. Burstification rate is the rate of injecting bursts into the OBS network. In this paper, a novel technique to control the burst loss ratio in OBS networks is proposed. The technique is based on classical control theory approaches to tune the burstification rate in order to achieve a desired burst loss ratio to satisfy the application requirements. Extensive simulations on the NSFNET topology show that the proposed technique achieves promising results. That is, the measured burst loss ratio hovers around the desired burst loss ratio for all nodes.

This paper considers the use of closed loop feedback control theoretic techniques to improve the performance of Optical Burst Switching (OBS) networks. In OBS networks, the Burst Loss Ratio (BLR) is the ratio between the lost bursts to the sent bursts. The BLR is used as a performance metric. The desired burst loss ratio depends on the application using the network. Some applications might tolerate more burst loss ratios than other applications. Higher network link utilization could be achieved by having more control over the burst loss ratio. Burstification rate is the rate of injecting bursts into the OBS network. In this paper, a novel technique to control the burst loss ratio in OBS networks is proposed. The technique is based on classical control theory approaches to tune the burstification rate in order to achieve a desired burst loss ratio to satisfy the application requirements. Extensive experiments show that the proposed technique achieves promising results. That is, the measured burst loss ratio hovers around the desired burst loss ratio and higher utilization is observed. Empirical approaches are used to identify the proposed model. The empirical model fits the OBS network by a value that did not fall below 75%.

Active Queue Management (AQM) policies are mechanisms for congestion avoidance, which pro-actively drop packets in order to provide an early congestion notification to the sources. Random Early Detection (RED), the defacto standard and its different flavors have been proposed as simple solutions to the AQM problem. However, these approaches require manual tuning and fail to accurately capture variations in the input traffic, thereby resulting in unstable behavior. α_SNFAQM is a new AQM mechanism that uses a neurofuzzy prediction method (α_SNF) to capture traffic variation and accurately detect the future congestion. It distinguishes (i) severe congestion and (ii) light congestion. We compare the performance of α_SNFAQM with other AQM schemes like RED, PAQM and APACE in a bottleneck link. Simulation results have shown that α_SNFAQM outperforms other AQM schemes in stabilizing the instantaneous queue length, reducing packet loss ratio while keeping a high utilization of the link.