The IEEE 802.11ad and IEEE 802.11ay standards define the medium access control (MAC) operation for millimeter-wave communications. During the beamforming training (BFT) period, the stations compete for the channel adopting a modified version of slotted aloha, where stations mandatorily transmit and can retry in a finite number of slots. This change makes the probability of a station accessing a given slot dependent on the previous ones. Additionally, stations halt the competition for BFT, if the number of consecutive collisions reaches a predefined collision threshold (CT) value. In this paper, we characterize the probability of a station interrupting the competition for BFT due to consecutive collisions. To address the characterization of the interruption for the BFT competition, we model the individual collision distribution and the distribution of a station reaching the CT over multiple BFT periods. The proposed model is validated through simulation and its performance is compared to other approximated IEEE 802.11ad models, confirming the accuracy of the proposed approach.

The range of applications that make up the Internet-of-Things ecosystem continues to grow. New opportunities present themselves along with new design challenges concerning the efficiency, portability, and processing capabilities of future Internet-of-Things systems. Thus, improving the operational metrics of individual Internet-of-Things devices, particularly across the edge and fog layers, is of paramount importance.In this manuscript, we present an approach for decreasing data collection at the edge, thus reducing form factor and power consumption of edge devices. This is particularly relevant for our application of interest, wearable motion capture, where human comfort and operational longevity are of prime importance. Our approach extrapolates from reduced edge data by leveraging prior physiological knowledge of the captured entity at the computational stage in the fog. By delegating computation to the fog, we also demonstrate the possibility for expanded capture volumes (operational areas) for future wearable motion capture systems and motion capture systems in general.Our approach, when prototyped on a millimeter wave sensor edge device and two fog node platforms of different processing tiers, shows that prior knowledge can facilitate a reduction in capture data dimensionality (and an associated decrease in power consumption) with little to no accuracy degradation, when compared to a more data-intensive edge system (Microsoft Kinect).

We analyze the consensus latency of Proof of Work (PoW) blockchains. A block reaches consensus if it is included in every miner’s longest chain. We define consensus latency in this paper as the duration from when a block is mined to when that block first reaches consensus. Understanding consensus latency helps us reason about when to confirm blocks for a series of confirmation rules. We find that blocks could reach consensus fast in practical scenarios. For instance, in a propagation network with maximum network delay delta, a block could reach consensus in about 1.3 deltas on average. Contrary to common expectations, average consensus latency even decreases when mining becomes easier. We establish an analytical model to understand the relation between consensus latency and various system parameters in various scenarios. We also build a blockchain simulator to validate the accuracy of our analysis. Our estimation closely tracks the simulation results.

Existing PoW blockchains adopt a simple k-block confirmation rule. A block is confirmed after it is buried sufficiently deep in the blockchain. The selection of the required depth k is largely based on empirical experiences and lacks a theoretical guarantee. We ask a fundamental question in PoW blockchains: when and how should miners confirm a block in PoW blockchains? We say a block is in the consensus state if it is included in every miner’s blockchain. The issue of the current k-block rule is that it does not depend on consensus. A block buried k-block deep may not be in the consensus state. We propose that a block should not be confirmed before the miner believes it to be in the consensus state. We present a consensus-inference protocol for miners to infer whether a block is in the consensus state locally. Based on the inference protocol, we further describe a new consensus-aware confirmation rule. Compared to the existing k-block rule, we show that the new rule achieves lower confirmation reversal probability under the same confirmation latency. The reduction could be orders of magnitude when multiple blocks are mined on average per network delay. Correspondingly, we further show that the new rule could improve the scalability of current blockchains by as much as 90% in common scenarios.

Existing PoW blockchains adopt a simple k-block confirmation rule. A block is confirmed after it is buried sufficiently deep in the blockchain. Such confirmation does not guarantee transaction finality. Specifically, after a block is confirmed, a miner may later confirm a different block in the same position of the blockchain. The previously confirmed block effectively gets reversed. A common reason for block reversal is that an adversary could privately maintain a malicious chain. The adversary releases the malicious chain once it is longer than the blockchain of at least one honest miner.In this paper, we propose a new confirmation method that relies on checkpoints to prevent the adversary from withholding its blocks. A checkpoint contains one or more blocks at the same position of the blockchain. Unlike existing protocols, we propose a simple mechanism to establish checkpoint blocks without introducing any trusted third parties. A committee is selected periodically to vote for checkpoint blocks. Miners only confirm a block when it is referenced by all checkpoint blocks at some block height, among other criteria. We prove the liveness and safety of this new confirmation rule. We show that conditioning on honest majority within a committee, miners achieve 0 confirmation reversal probability. We demonstrate that with majority mining power being honest, we could always select a proper checkpoint period to ensure the assumption of the majority of a committee being honest holds for hundreds of years on average. In the rare case where this assumption is broken, miners in the checkpoint-enabled blockchain still incur the same reversal probability as in the current k-block confirmation.

The detection of abnormal sequences of SIP messages in real-time is crucial to avoid SIP signaling-based attacks. In this paper, we propose a deep learning approach to detect signaling patterns of multimedia sessions established with the Session Initiation Protocol (SIP). The approach is based on a recurrent neural network (RNN). We study the performance of different Long Short-term Memory (LSTM) RNN architectures, which are trained using a SIP signaling dataset of trustworthy SIP dialogs captured by a SIP server. The trained RNNs are then used to detect the SIP dialogs in real-time. After characterizing the dataset adopted for the training, validation, and testing, we present the experimental results obtained for the different RNN architectures, showing that the classification probability of trustworthy SIP dialogs exceeds 93% in the test stage. Finally, we present two methodologies to detect abnormal SIP dialogs, i.e., not contained in the trustworthy training dataset. After a detailed analysis of the skewness and kurtosis computed with the numerical RNN outputs, we show that they can be used as classification features. The first method is based on a K-means unsupervised classifier, while the second one is based on a semi-supervised threshold-based classifier. Experimental results show that the threshold-based classifier achieves 99.45% of detection probability, showing the effective utility of the proposed methodology to detect abnormal SIP sequences in a short period of time.

More and more complex applications are deployed in virtualized Cloud environments. Cloud applications usually consist of multiple virtual components and the data flow among these components tends to be highly complex and unpredictable. Cloud providers have no visibility into tenants' virtual components and rely on non-intrusive knowledge to infer the service performance. The scale, complexity, and heterogeneity make Cloud environments prone to different kinds of performance anomalies. An automatic anomaly detection mechanism is necessary to help avoid service failure at an early stage of anomalies. We propose DMFD, a decentralized and non-intrusive anomaly detection system for Cloud applications. DMFD considers each virtual component as a black box, and works without knowledge from underlying operating system or applications. DMFD infers the dependency model to characterize the complex interaction behavior among multiple components, and further builds an adaptive data flow ratio model for anomaly detection. The data flow model explores a simple and common relationship that the number of outgoing response data packets changes accordingly with the number of incoming request data packets at each component. We evaluate DMFD with several different kinds of Cloud applications in our data center. The results reveal that DMFD has high precision and low false positive/negative in detecting different kinds of performance anomalies under varying workload scenarios.

Cloud vendors offer content delivery network (CDN) services to compete for the video market. The user experience and the costs of providing the same video streaming service can vary when using different cloud CDNs. We emulate video streaming users in PlanetLab cloud to measure cloud CDNs including Amazon Web Service (AWS) CloudFront, Microsoft Azure Verizon CDN, and Google Cloud CDN. We leverage an approximated Quality of Experience (QoE) as a metric for evaluation. Our study finds that: 1) cloud vendors vary in providing QoE across regions; the video provider should assign a user to the CDN offering the best QoE at his location; 2) the QoE provided by one CDN can change over time; the video provider should adapt the CDN selection according to the real time QoE measurement; 3) cloud CDNs vary in scalability; streaming sessions may crash when there is bursty user demand; video providers should choose among the cloud CDNs that can properly scale; 4) regarding the cost, some cloud CDN is more economical than others given certain cache hit rate; video providers can minimize their costs by forcing free trial users to stream from the cheapest one.

Cloud applications run on numerous servers and network elements. Complexity and heterogeneity of these elements make evaluation of application performance very challenging without invasive application-specific probe. We propose an anomaly detection system for Cloud application servers. Our system only collects ingress and egress data packet counts. We observe that when an application server is operating normally, the egress data packet count is highly correlated to the ingress data packet count. When the server is in an abnormal state, the correlation between ingress and egress packet counts decreases, regardless of the cause of the anomaly. This observation is the basis of our system. We validate our observation and the proposed system using heterogonous benchmarks with heavy tail service distribution and ten different anomaly types in the Cloud. We demonstrate that our system can accurately detect all injected anomalies with 1.6% false positives and 0.23% CPU overhead.

Quality of Experience (QoE) anomalies widely exist in all types of video services. As video services migrate to the Cloud, unique challenges occur to deploy video services in the Cloud environment. We study the QoE anomalies for users in a video service deployed in a production Cloud CDN. We use a QoE anomaly identification system, QRank, to identify anomalous systems. We consider Cloud CDN servers, Cloud CDN networks, transit networks, user access networks and different types of user devices. Our extensive experiments in production Cloud find several interesting insights about QoE anomalies of video streaming in the Cloud. 91.4% of QoE anomalies are detected on 15.32% of users. These users experience QoE anomalies persistently and recurrently. The Cloud servers and networks seldom cause QoE anomalies. More than 99.98% of QoE anomalies are identified in anomalous systems including the transit networks, the access networks and user devices. We infer that transit networks are the actual bottleneck systems for QoE anomalies in production Cloud. More than 95% of persistent and recurrent QoE anomalies are identified in less than 10 transit networks. We collect latency measurements to anomalous networks and the analysis indicates that the limited capacity in transit networks are the major cause of QoE anomalies. Resulting anomalies impair user QoEs persistently or recurrently. In order to provide good user QoE, the Cloud provider should identify transit networks that may become bottlenecks for high quality video streaming and appropriate peering with Internet Service Providers (ISPs) to bypass these bottlenecks.

Infrastructure-as-a-Service environments are becoming increasingly popular. When there is a failure, many applications require service restoration within a few seconds. Reaction to failures in Cloud is still slow for many applications. Monitoring is limited to instance metrics that are not conducive to precise diagnosis due to complexity of virtualization in physical hosts. Interferences among different VMs complicates the diagnosis. We propose a new dynamic monitoring module as a part of multi-agent based cloud management framework named LAMA. Applications are at the center of our framework. Agents distributed throughout the cloud infrastructure are responsible for aggregating metrics and performing customized diagnostics for each application. This reduces overhead, removes centralized bottlenecks and allows customized configuration with finer granularities. Our approach is also environment aware as each app agent has access to the application, virtual instances and hosting infrastructure metrics. This feature enables the creation of more efficient diagnostics algorithms customized to application's needs. We develop and deploy LAMA in our datacenter to demonstrate (1) how an integrated approach with access to the state of the app's improves efficiency of failure detection, (2) how our monitoring and diagnosing architecture can improve load distribution in the network and (3) the impact of finer granularity on failure detection time.

Commercial large-scale VoD systems such as Netflix and Hulu rely on CDNs to deliver videos to users around the world. Various anomalies occur often and degrade users' Quality of Experience (QoE). Detecting and locating such anomalies are highly complex due to a large number of different entities involved in the end-to-end video delivery. These entities include VoD provider, CDN/Cloud providers, transit ISPs, access ISPs, and end user devices. QoE perceived by the users is a critical metric for VoD providers. We propose QWatch, a scalable monitoring system, which detects and locates anomalies based on the end user QoE in real-time. We evaluate QWatch in a controlled VoD system and production Microsoft Azure Cloud and CDN. QWatch effectively detects and locates QoE anomalies in our extensive experiments. We discuss insights obtained from running VoD system with 200 worldwide users in production Cloud.

We are increasingly reliant on cellular data services for many types of day-to-day activities, from hailing a cab, to searching for nearby restaurants. Geo-location has become a ubiquitous feature that underpins the functionality of such applications. Network operators can also benefit from accurate mobile terminal localization in order to quickly detect and identify location-related network performance issues, such as coverage holes and congestion, based on mobile measurements. Current implementations of mobile localization on the wildly-popular Android platform depend on either the Global Positioning System (GPS), Android's Network Location Provider (NLP), or a combination of both. In this paper, we extensively study the performance of such systems, in terms of its localization accuracy. We show through real-world measurements that the performance of GPS+NLP is heavily dependent on the mobility of the user, and its gains on localization performance is minimal, and often even detrimental, especially for network round-trip delays up to 1s. Building upon these findings, we evaluate the efficacy of using Tattle, a cooperative local measurement-exchange system, and propose Delay-Adjusted U-CURE, a clustering algorithm that greatly improves the localization performance of both GPS-only, and GPS+NLP techniques, without keeping expensive system states, nor requiring any location anchors nor additional instrumentation, nor any external knowledge that is not available programmatically to application designers. Our results are promising, demonstrating that median location accuracy improvements of over 30% is achievable with just 3 co-located devices, and close to 60% with just 6 co-located devices. These findings can be used by operators to better manage their networks, or by application designers to improve their location-based services.

Cellular service subscribers are increasingly reliant on cellular data services for all kinds of mobile applications. Oftentimes, when subscribers experience frustratingly high network delays and timeouts, they like to know whether their experiences are shared by other users nearby. The question that is often asked is essentially this: “is it just me, or do others around me face the same problem?” In this paper, we describe how we use Tattle, a distributed real-time participatory sensing and monitoring framework, to glean network performance information from users nearby. Tattle relies on recent advances in peer-to-peer device networking, such as Wi-Fi Direct, Bluetooth Low Energy, and Apple's iBeacon, to exchange key snippets of diagnostic information using very low-power, very short-range local-area wireless interfaces, between participating devices. We propose and develop a robust statistical algorithm, based on quantile regression, which identifies key points in time where a device experiences high delays and outages that are not observed by its neighbors, and decides if the device is performing “normally”, or “abnormally”. This directly answers the “me, or others?” question. We demonstrate and validate the efficacy of our system through real-world measurements of network delay, consisting of over 7,300 time-series that comprises over 443,500 data samples, using commodity smart devices attached to two different providers' networks.

Detection of anomalies in large Cloud infrastructure is challenging. Understanding operational behavior of Cloud is extremely difficult due to the heterogeneity of different technologies, virtualized platforms and complex interactions among the systems. Many of existing system models for Cloud are based on utilization metrics such as CPU, memory, network and I/O. Such system models are quite complex and their anomaly detection mechanisms are mostly based on threshold scheme. Utilization metrics exceeding a certain threshold would trigger an alarm. In fact, it is impossible to determine proper threshold for all anomalies. These system models fail to assess the state of the system accurately. We propose a novel anomaly detection system based on user perception rather than complex system models. In our Perception-Based Anomaly Detection system (PBAD), each component within multi-tier applications monitors response time and determines whether overall service response time is adequate. PBAD also locates the anomaly by analyzing component behaviors. PBAD masks the complexity of Cloud and addresses what matters, how user perceives the service provided by the Cloud applications. The key advantages of the proposed algorithm are simplicity and scalability. We implement and deploy PBAD in our production data center environment. The experimental results show that PBAD detects numerous types of anomalies as well as the combination of anomalies where existing systems fail.

The automatic discovery of dependencies in distributed Cloud applications is very useful for large scale deployments. Dependencies can be used to identify the anomalies due to errors, failures or the performance bottleneck in applications. Although existing dependency models can be useful, we believe more comprehensive dependency model would improve anomaly detection in large scale distributed applications. We propose a VM dependency discovery system and introduce dependency primitives that incorporate complex application behavior/interaction patterns. We also formulate response time characteristics for each dependency primitive. Using the component dependencies and traffic monitoring, we develop a stochastic model to estimate the response time probability distribution for components and overall application. We evaluate and validate our system with various production applications. Experiments show that we can accurately discover application dependencies and also predict not only the average response time but the 95th percentile response time within 8% of the actual response time.

In commercial Video-on-Demand (VoD) systems, user's Quality of Experience (QoE) is the key factor for user satisfaction. In order to improve user's QoE, VoD providers replicate popular videos in geo-distributed Cloud and deploy cache servers close to users. Generally, the VoD provider selects a server for the user request according to the user's location. Usually geographically closely located servers would provide lower network delay. However, the performance of VoD servers deployed in cloud virtual machines (VM) depends not only on the network delay but also resource contention due to other VMs and highly dynamic user demands. Thus, QoE offered by the server varies greatly over time as user demands and network traffic fluctuate regardless of the location. Selecting a server close to users sometimes reduces the network delay but cannot guarantee QoE in general. We believe that end users have the best perception of server performance in terms of their QoE rather than the servers themselves. What user perceives incorporate performance of all elements, such as network delay and server response time in VoD service. We propose VoD server selection schemes that dynamically select servers according to user's QoE feedback. We integrate our server selection schemes with Dynamic Adaptive Streaming over HTTP (DASH) clients and evaluate our system both in simulation and in Google Cloud. Results show our system improves user QoE up to 20% compared to existing solutions.

Fast detection of performance anomalies is critical in Cloud applications, but challenging to implement in a general and effective tool with low operational overload. We propose FSAD, a performance anomaly detection system based on the concept of flow similarity. It stems from the observation that, in general, the number of responses generated by a component closely follows the number of received requests, but this relation stops holding in presence of performance anomalies. In FSAD, components are regarded as black boxes, and time series of incoming and outgoing packets are fed to the flow similarity analysis for anomaly detection. The effectiveness of FSAD is demonstrated in experimental results.

Many complex application services are deployed in virtualized Cloud environments. Cloud applications consist of multiple components and the data flow among these components tends to be highly complex and unpredictable. The complexity and heterogeneity make anomaly detection challenging. We propose FlowBox, a distributed anomaly detection system for Cloud applications. FlowBox considers each server component as a black box and detects performance anomalies using the flow analysis. The black box model addresses the challenge of accurately describing the complex system model. The flow analysis is based on a simple relationship of data flow in any given component of Cloud applications. Between any two components, the number of requests should always be equal to the number of responses within a given time interval during normal operations. FlowBox monitors traffic flow in each component and continually builds flow signatures in order to describe the normal application behavior. Using the flow signatures, FlowBox detects performance anomalies in Cloud applications. We evaluate FlowBox with several different kinds of Cloud applications in our datacenter. Experimental results show that FlowBox achieves 96.02% detection precision, 3.98% false positive, and 3.5% false negative in detecting various kinds of anomalies.

As VoD systems migrate to the Cloud, new challenges emerge in managing user Quality-of- Experience (QoE). The complexity of the cloud system due to virtualization and resource sharing complicates the QoE management. Operational failures in the Cloud could be challenging for QoE as well. We believe that end users have the best perception of system performance in terms of their QoE. We propose a QoE based adaptive control system for VoD in the Cloud. The system learns server performance from the user QoE and then adaptively selects servers for users accordingly. We deploy our proposed system in Google Cloud and evaluate it with hundreds of clients deployed all over the world. Results show that given the same amount of resources, our system provides 9% to 30% more users with QoE above the Mean Opinion Score (MOS) "good" level than the existing measurement based server selection systems. The system guarantees a better QoE (above 6% better) for 90% users. Additionally, our system discovers operational failures by monitoring QoE and prevents streaming session crashes. A computational overhead analysis shows that our system can easily scale to large VoD systems containing thousands of servers.

The operating environment of cellular networks can be in a constant state of change. One Singaporean operator expressed difficulty with the coverage assertion (CA) problem of whether regulated minimum coverage is met, especially in urban areas. Currently, the operator manually appraises coverage through laborious and expensive walk/drive-tests. In this paper, we propose Tattle, a distributed, low-cost and comprehensive cellular network measurement collection and processing framework. We exemplify Tattle by leveraging on participating UEs to report on network coverage in real-time. Tattle exploits wireless local-area interfaces to exchange RSCP measurements amongst devices to preserve the co-locality of readings and conserve power. We propose U-CURE, a clustering algorithm which considers sample location uncertainty and the knowledge of device co-location to remove erroneously localized readings. We develop a prototype app on the Android™ platform as a proof-of-concept of the Tattle framework. We then use the Tattle framework to perform extensive RSCP measurement collection and processing in various areas in Singapore, collecting over 3.78 million readings. We present visualizations of mean signal coverage and RSCP CDFs for various areas of interest. The latter is a key output of Tattle, which helps operators to appraise coverage and solve the CA problem by relying on subscriber measurements, instead of expensive, laborious and limited-scale walk-/drive-tests.

Multi-tenant data centers host a high diversity of applications with continuously changing demands. Applications require response times ranging from a few microseconds to seconds. Therefore, network traffic within the data center needs to be managed in order to meet the requested SLAs. Current feedback congestion control protocols may be too slow to converge to a stable state under high congestion situations. Sudden bursts of traffic from heterogeneous sources may render any reactive control inefficient. In this paper, we propose PACE, a preventive explicit allocation congestion control protocol that controls resource allocations dynamically and efficiently. PACE specifically addresses Data Center requirements: efficient network usage, flow completion time guarantees, fairness in resource allocation, and scalability to hundreds of concurrent flows. PACE provides micro-allocation of network resources within dynamic periods, lossless communication, fine-grained prioritization of flows, and fast adaptation of allocations to the arrival of new flows. We simulate PACE and compare it with recent proposed protocols specially addressed to Data Centers. We demonstrate that PACE is fairer, in particular for short flows, flows with different RTTs and a higher number of concurrent flows. It also maintains high efficiency and controlled queue usage when exposed to sudden bursts.

Virtualization allows us to consolidate multiple servers onto a single physical machine, saving infrastructure cost. Yet, consolidation can lead to performance degradation, jeopardizing Service Level Agreement (SLA). In this paper, we analyze and identify the factors to the performance degradation due to consolidation - that is the wait time and the ready time. The wait time is the queuing time caused by other virtual machines (VMs). The ready time is the time the resource takes to be ready to service, such as the seek time incurred in traditional storage. The ready time can substantially deteriorate the request response time. Unfortunately, existing schedulers can only manage the wait time, but not the ready time. To control both quantities, we propose an adaptive disk scheduler called DPack. DPack schedules the VMs based on the likelihood of the VM failing the SLAs. DPack then adjusts the exclusive access time based on the VM resource access prediction. DPack considers the workload changes and request arrival to enhance robustness. We develop DPack based on the default disk scheduler in KVM and evaluate it against several existing disk schedulers available in KVM and Xen. The results show that DPack can improve the 99th percentile response time up to 76%. In the highly consolidated environment, DPack can also satisfy all the SLAs, while the other schedulers cannot meet the SLAs for at least 50% of the VMs.

The adequate operation of IPTV distribution networks heavily relies on the effective maintenance and management of their underlay DSL infrastructure. New hardware and software is required in order to improve monitoring capabilities and to directly diagnose anomalies that other segments of the DSL network cannot identify. In this work we initially compare the accuracy performance of SVM-specific formulations for constructing a robust ground truth within our classification procedure regarding abnormalities issued at anomaly-aware Digital Subscriber Line Access Multiplexers (DSLAMs) of the DSL infrastructure. Moreover, we consider the pragmatic cost of repairing anomalies that were misclassified and characterize each classifier according to the overall cost that is possible to incur to the network operator. In parallel, this work attempts to practically improve the network-wide anomaly classification performance by proposing a semi-supervised classification scheme that updates the initial supervised scheme by testing unlabelled anomalies occurring at anomaly-unaware DSLAMs.

Cloud computing environments provide on-demand resource provisioning, allowing applications to elastically scale. However, application benchmarks currently being used to test cloud management systems are not designed for this purpose. This results in resource underprovisioning and quality-of-service (QoS) violations when systems tested using these benchmarks are deployed in production environments. We present C-MART, a benchmark designed to emulate a modern web application running in a cloud computing environment. It is designed using the cloud computing paradigm of elastic scalability at every application tier and utilizes modern web-based technologies such as HTML5, AJAX, jQuery, and SQLite. C-MART consists of a web application, client emulator, deployment server, and scaling API. The deployment server automatically deploys and configures the test environment in orders of magnitude less time than current benchmarks. The scaling API allows users to define and provision their own customized datacenter. The client emulator generates the web workload for the application by emulating complex and varied client behaviors, including decisions based on page content and prior history. We show that C-MART can detect problems in management systems that previous benchmarks fail to identify, such as an increase from 4.4 to 50 percent error in predicting server CPU utilization and resource underprovisioning in 22 percent of QoS measurements.

Consolidating Virtual Machines (VMs) in data centers is desirable as it reduces hardware and power costs. However the performances of VMs on shared physical servers are not isolated from each other as they contend for the same server resources. This contention degrades the performance of delay sensitive applications and can increase response times by three orders of magnitude at high contention levels. In order to achieve Service Level Agreements (SLAs) under VM consolidation, resources must be allocated by considering the performance effects of contention. We therefore present VARACO, a contention-aware VM management system to achieve Quality-of-Service (QoS) targets for multi-tier web applications. VARACO models applications' performances online using 13 server resource utilization and contention metrics. Resources are dynamically allocated using these models to achieve QoS targets. Our results show that application- level performance can be modeled 130% more accurately when resource contention is considered. We demonstrate VARACO by achieving a 90thpercentile response time target of a sample application under VM consolidation.

We expect femto-cells to be massively and densely deployed in the future. Numerous existing works on femto-cell interference management assume that the local topology of interfering femto-cells can be sufficiently approximated through sensing, if not already known in advance. We show that this assumption results in poor throughput performance in dense femto-cell networks. For some cell-edge users, using conventional sensing in dense deployments can result in almost 50 times less instantaneous throughput, as compared to having oracular knowledge of interference topology. This sub-optimality is caused by “far-away” interferers. These are femto-cells that are deployed just far enough such that their presence will not be detected by conventional sensing. We then introduce a mobile sensing scheme to detect these “far-away” interferers by exploiting the inherent mobility of femto-cell users. We show through packet-level simulation that this sensing scheme is able to better approximate the interference topology. This results in significantly improved performance over conventional sensing, in dense deployment scenarios.

IPTV networks blindly rely on the adequate operation and management of the underlying infrastructure that in numerous cases is threaten by unexpected anomalous events which consequently cause QoS degradation to the end-user. Thus, it is of great importance to deploy techniques embodied with diagnostic and self-protection metrics for determining and predicting the arrival of such events in order to proactively charge defense mechanisms without the need of an exhaustive manual inspection by the network operator. In this paper we propose and demonstrate the applicability of the Rényi entropy as a useful diagnosis feature for explicitly characterizing DSL-level anomalies issued in an IPTV network of a large European ISP. It is revealed that different orders of the Rényi entropy can formulate meaningful detection and categorization of phenomena occurring on specific Digital Subscriber Line Access Multiplexers (DSLAMs) within the DSL infrastructure. Via the synergistic exploitation of the local maxima peaks generated by each Rényi-based distribution we exhibit the feasibility to extract and identify lightweight anomalies that under simple metrics cannot be detected.

Maximizing consolidation ratio, the number of virtual machines (VMs) in a physical machine, without violating customers' SLAs is an important goal in the cloud. We show that it is difficult to achieve this goal with existing hypervisor schedulers. The schedulers control only the amount of resource allocation, but not the sequence of VM execution. This sequence can significantly impact the response time when requests arrive concurrently for the VMs sharing the same CPU. We find that the response time can increase as much as 100% for every additional VM in the system, even if the utilization does not exceed the maximum capacity. Therefore, existing schedulers have to reduce the consolidation ratio to meet SLAs. Previous resource-provisioning works rely on existing schedulers that cannot guarantee SLAs without reducing the consolidation ratio. We propose SageShift, a system that can achieve SLAs without penalizing the consolidation ratio. SageShift consists of a VM admission control - Sage, and a hypervisor scheduler - Shift. To admit a VM, Sage assesses feasibility of its SLA based on the patterns of incoming requests. Shift maintains the admitted SLAs by adjusting both the amount of resource allocation and the sequence of VM execution. The dynamic adjustment is based on the observed response time and the SLAs. We modify the KVM scheduler in Linux kernel to implement Shift. We show that Shift can improve the consolidation ratio by 66% without compromising the SLAs. Under bursty incoming requests, Shift maintains all SLAs within 3% of the percentile target. But existing schedulers in VMware ESXi, Xen and KVM fail to meet one or more SLAs with up to 33% below the percentile target. Shift is also work-conserving. It allows best-effort VMs to run in the background in order to maximize hardware utilization without impacting SLAs.

Symmetric multiprocessing (SMP) virtual machines (VMs) allow users to take advantage of a multiprocessor infrastructure. Despite the advantage, SMP VMs can cause synchronization latency to increase significantly, depending on task scheduling. In this paper, we show that even if a SMP VM runs non-concurrent applications, the synchronization latency problem can still occur due to synchronization in the VM kernel.

Next-generation wireless cellular networks are likely to be multi-tiered, consisting of a macro-cell tier based on the traditional operator-deployed macro-cell structure, and a lower tier of femto-cells that are arbitrarily deployed by end-users. The need for such architectures is driven by increasing demands for higher data rates as well as to provide capacity to serve increasing numbers of subscribers. This paradigm shift in cellular architecture may therefore imply changes in the way cellular networks are managed today. In this paper, we examine operation costs of a two-tier macro-femto-cellular network in a capacity-limited scenario. We identify the key parameters, including nomadic-to-mobile user ratio, non-linear femto-cell capacity growth, femto-cell subscription rate, and QoS provisioning costs, which will impact recurring annual management costs of these networks. We then identify a model which captures these cost components and key parameters, allowing an operator to characterize the management cost of their access networks. We also identify through evaluations key situations where adopting these femto-cells may in fact add to an operator's overall costs, and provide key insights into the financial feasibility of femto-cells.

Network operators in large-scale networks are often faced with long lists of maintenance tasks and find it difficult to track the relative importance of these tasks, without knowing their impact on the network's operation. As a result, operators may react slowly to critical tasks, increasing network downtime and maintenance costs. We present a system that quantifies the impact of maintenance tasks so that operators can prioritize their reaction according to the estimated impact (i.e., spend more time and effort on avoiding the disruption caused by high-impact maintenance tasks). In particular, the proposed system estimates the amount of traffic loss due to maintenance operations on inter-domain routing sessions, one of the most frequently modified aspects of network configurations. We implement the proposed system and apply it to 372 routing sessions in a nation-wide ISP network. The system identifies sessions with a varying degree of impact: sessions with nearly zero data loss, as well as sessions that can result in more than 1,000 GB of data loss if disrupted without any protection mechanism applied. We also show that predicting the amount of data loss is not straightforward since this amount changes over time, often in unexpected ways (e.g., from 50GB to 0 over one-month period). Therefore, the proposed impact analysis system is necessary for network operators to perform periodic audits of the routing sessions' impact and to classify the sessions according to the projected data losses. Operators can then decide the level of protection for each session (e.g., employ more effective and costly methods to protect critical sessions) and thus allocate maintenance costs more efficiently.

Network configurations implement a set of policies that control a network's behavior. Therefore, correct understanding of the configurations is vital to ensure that the network operates according to the intended policies. However, the current practice of manually reading a large number of configuration commands, which are written in low-level languages and distributed in multiple devices, is inefficient and significantly increases management costs and operator errors. We propose a system that helps decode network configurations by interpreting low-level fragmented configurations and then presenting their high-level intended policies. In particular, the proposed system is applicable to inter-domain routing policies, one of the most complex aspects of network configurations. We implement our system and evaluate its effectiveness through a set of user studies involving 44 participants. These studies examine the participants¿ comprehension of routing policies presented with our system as compared to those presented with existing configuration languages. The studies show that our system improves both accuracy, from 70% to nearly 100%, as well as time-to-task-completion, from 30 minutes to 10 minutes. We believe that our system provides a basis for a clean separation of policy intent from its implementation so that policies can be better designed and understood. We also discuss the weaknesses in usability of current network configurations and argue that all aspects of future management systems need to be designed to address these usability issues.

Configuring a network is a tedious and error-prone task. In particular, configuring routing policies for a network is complex as it involves subtle dependencies in multiple routers across the network. Misconfigurations are common and certain misconfigurations can bring the Internet down. In 2005, a misconfigured router in AS 9121 blackholed traffic for tens of thousands of networks in the Internet. This paper describes NetPiler, a system that detects router misconfigurations. NetPiler consists of a routing policy configuration model and a misconfiguration detection algorithm. The model is applicable to routing policies configured on a single router as well as to network-wide configuration. Using the model, NetPiler detects configuration commands that do not influence the behavior of the network - we call these configurations ineffective commands. Although the ineffective commands could be benign, sometimes when the commands are mistakenly configured to be ineffective, they cause the network to misbehave deviating from the intended behavior. We have implemented NetPiler in approximately 128,000 lines of C++ code, and evaluated it on the configurations of four production networks. NetPiler discovers nearly a hundred ineffective commands. Some of these misconfigurations can result in loss of connectivity, access to protected networks, and financial implications by providing free transit services. We believe NetPiler can help networks to significantly reduce misconfigurations.

As a network evolves over time, multiple operators modify its configuration, without fully considering what has previously been done. Similar policies are defined more than once, and policies that become obsolete after a transition are left in the configuration. As a result, the network configuration becomes complicated and disorganized, escalating maintenance costs and operator faults. We present a method called NetPiler, which groups common policies by discovering a set of shared features and which uses the groupings for the configuration instead of using each individual policy. Such an approach removes redundancies and simplifies the configuration while preserving the intended behavior of the configuration. We apply NetPiler to the routing policy configurations from four different networks, and reduce more than 50% of BGP communities and the related commands. In addition, we show that the reduced community definitions are sufficient to satisfy changes as the network evolves over nearly two years.

We present a new switch fabric architecture on a chip for the next generation link speed. The advances in CMOS technologies and optical technologies have changed the design constraints of switch fabrics. The number of transistors on a chip is no longer a major issue in the switch fabric design. Processing speed and wire length have become critical as the link rate continues to increase dramatically. Noting these trends, we propose a new switch fabric that consists of switching elements in a regular tile structure. This regular structure leads to a highly pipelined and self-routing architecture. The results of the synthesized circuits show that a single chip can support 64 ports and 40 Gb/s data rate using current CMOS technologies.

Interdomain routing is the glue that holds together the disparate networks of the Internet. However, scalability constraints force individual networks to only have a local view of the Internet's routing state. This lack of global view hinders the troubleshooting of Internet routing problems and limits the routing choices of network operators. In this paper, we propose a p2p-based infrastructure for the Internet-wide management of interdomain routing. Peers dispersed in various Internet sites collect information from the participating networks and exchange knowledge in the event of a routing problem. Thus, the infrastructure as a whole carries a global routing view. Participating networks use this infrastructure in order to obtain an insight on remote routing events and potentially respond accordingly. Our approach is novel in that it exploits the peer-to-peer paradigm for routing management purposes. The proposed infrastructure is resilient in terms of information availability, self-organizing, lightweight, and easily deployable. We illustrate the need, design, and benefits of such a management infrastructure.

Router configuration is a difficult and complex task. At the same time, it is a crucial task as it accounts for a network's profit, performance, and security. Routing policies are configured in low-level languages and the high- level intent is hard to decipher. In this paper, we propose a system, called NetPolis, which abstracts the high-level intents from low-level configuration of routing policies. The goal of NetPolis is to automatically generate the inter- domain routing policy configuration of a network. NetPolis takes the network's router configuration files and compares the import and export policies from various perspectives and granularities. The output is a multi-level model to represent neighbor networks with similar routing policy enforcements. We validate our approach by applying NetPolis to the router configuration files from a production network of a major ISP. The network operator confirms that NetPolis provides a compact summary of the network's routing policies and this summary helps to verify high-level intents, to identify misconfigurations, and to aid in policy modifications.

Configuring a network is a low-level, device-specific task. Many have compared it to writing a distributed program in assembly language, reserved only for highly experienced network operators. Automation has been proposed by researchers and industry as the solution to problems in network configuration. However, there is a certain amount of resistance from the operator community against automation. On the one hand, operators do desire a way for network-wide configuration. On the other hand, they still like to have access and control to details, to ensure flexibility and for debugging. In this paper, we attempt to answer the question "How should we automate network configuration" by studying where the complexity lies in network configuration. With an operational perspective, using data from three different types of production networks, we analyze the configuration files from these networks over the span of up to two years. Our analysis shows that the majority of changes to these files are a few lines each and made frequently. We found that routing, especially its policies, constitute a significant portion of the configuration files, as well as modifications to them. We then present complexity models to measure network-wide risk, impact and duplication of routing policies in network configuration. We show that risk and impact tend to grow over time, and the duplication factor is high. Based on the results of our analysis, we propose ways to automate the complex parts of network configuration.

Entropy-based approaches for anomaly detection are appealing since they provide more fine-grained insights than traditional traffic volume analysis. While previous work has demonstrated the benefits of entropy-based anomaly detection, there has been little effort to comprehensively understand the detection power of using entropy-based analysis of multiple traffic distributions in conjunction with each other. We consider two classes of distributions: flow-header features (IP addresses, ports, and flow-sizes), and behavioral features (degree distributions measuring the number of distinct destination/source IPs that each host communicates with). We observe that the timeseries of entropy values of the address and port distributions are strongly correlated with each other and provide very similar anomaly detection capabilities. The behavioral and flow size distributions are less correlated and detect incidents that do not show up as anomalies in the port and address distributions. Further analysis using synthetically generated anomalies also suggests that the port and address distributions have limited utility in detecting scan and bandwidth flood anomalies. Based on our analysis, we discuss important implications for entropy-based anomaly detection.

In this paper, we analyze network traffic behavior by decomposing header traffic into control and data planes to study the relationship between the two planes. By computing the cross-correlation between the control and data traffics, we observe a general ‘similar’ behavior between the two planes during normal behavior, and that this similarity is affected during abnormal behaviors. This allows us to focus on abnormal changes in network traffic behavior. We test our approach on the Network Intrusion Dataset provided by the Information Exploration Shootout (IES) project and the 1999 DARPA Intrusion detection Evaluation Dataset from the MIT Lincoln Lab. We find that TCP control and data traffic have high correlation levels during benign normal applications. This correlation is reduced when attacks that affect the aggregate traffic are present in the two datasets.

A trend in the current network industry is to use the dominant Ethernet technology for large-scale networks due to its cost effectiveness and simple management. In this paper, we address scalability and configuration issues in large-scale Ethernet networks. Specifically, a hierarchical architecture and a self- configuration process of network devices are proposed. New address allocation and distribution schemes are introduced to solve the scalability issue of the traditional Ethernet networks. The proposed architecture is interoperable with traditional protocols of hosts. Therefore, this architecture can be deployed without modifying end hosts' protocols.

Routing infrastructure plays a vital mle in the Internet, and attacks on routers can be damaging. Compromised routers can drop, modih, mislforward or reorder valid packets. Existing proposals for secure forwarding require substantial computational overhead and additional capabilities at routers. We propose Secure Split Assignment Trajectory Sampling (SATS), a system that detects malicious routers on the data plane. SATS locates a set of suspicious routers when packets do not follow their predicted paths. It works with a traffic measurement platform using packet sampling, has low overhead on routers and is applicable to high-speed networks. Different subsets ofpackets are sampled over dzyerent groups of routers to ensure that an attacker cannot completely evade detection. Our evaluation shows that SATS can signzjicantly limit a malicious router's harm to a small portion of traffic in a network.

Recently, Ethernet technology has received a lot of attention for large area networks as well as local area networks (LANs) due to its economical advantage. However, the Ethernet technology has scalability issues when it is applied to large area networks. In this paper, we present hierarchical broadcast ring architecture that provides scalability to high-speed Ethernet networks. The proposed Ethernet switch-based network architecture adopts mechanisms of IP router-based networks. The logical ring topology of the architecture provides the broadcast service without a broadcast storm problem. A shortest path link-state routing is used for data forwarding for high link utilization. Furthermore, the hierarchical proxy ARP (Address Resolution Protocol) scheme and an aggregation scheme of IP addresses are proposed to reduce the number of broadcast frames and the forwarding table size. As a result, the hierarchical broadcast ring architecture can support millions of hosts with a high-speed Ethernet network. The proposed architecture maintains the simple and efficient management and operation of the Ethernet network for large area networks.

In cellular networks, an important call-level quality-of-service (QoS) issue is how to limit the probability of forced termination during handoffs. One solution is to predict the trajectory of mobile terminals so as to perform bandwidth reservation in advance. With the vision that future mobile devices are likely equipped with reasonably accurate positioning capability, we propose a novel mobility prediction technique that incorporates both mobile positioning information and road topology knowledge. We then develop an adaptive bandwidth reservation scheme that dynamically adjusts the reservation at each base station according to both incoming and outgoing hand-off predictions generated using our mobility prediction technique. We evaluate the performance of the scheme via simulations, along with six other schemes for comparison purposes. Results agree with intuition that schemes which incorporate more knowledge are able to achieve better reservation efficiency. Our scheme is shown to achieve the best efficiency among all realizable schemes simulated.

We develop integrated algorithms for dynamic provisioning of SLA-guaranteed IP services in IP-over-optical networks. By integrated, we mean that the provisioning mechanisms takes into account topology, resource, and protection information in both the IP and optical layers. This is achieved through two novel path selection algorithms developed in this paper, and through the GMPLS common control plane. The performance study shows that our proposed algorithms can really provide guarantee across multiple levels of service. In addition, we can achieve significant improvement over existing integrated provisioning approaches in terms of bandwidth utilization, request blocking probability, and algorithm complexity.

In the past, wireless network reliability issues have received limited attention. Previously proposed solutions have focused on protection against base station failures, which may require significantly more base stations than necessary. Actual data from wireless carriers suggest that problems in the backhaul are the uttermost reliability problems faced, and is therefore the main focus of this work. Current radio access networks are based on tree and star-like topologies, which have no inherent restorability properties. We propose a heuristic topology enhancement method that adds redundant spans and upgrades existing infrastructure cost-effectively, in order to create partially meshed architectures that could provide the desired level of restorability against single span failure scenarios. Both span and path restoration techniques are explored. The algorithm was tested using several different variants of restoration mechanisms. Results show that the proposed heuristic algorithm is able to achieve reasonably good solutions in a time scale that is several orders of magnitude faster than an optimization approach based on binary integer programming formulation.

In future wireless broadband networks, bandwidth demands could fluctuate abruptly due to movement of high data rate users. Dynamic bandwidth reservation plays a key rule in ensuring that mobile calls are not disrupted after they are admitted into the network. We propose a dynamic guard bandwidth scheme that adapts the amount of guard bandwidth in both wired and wireless links according to real-time handoff predictions made by individual mobile terminals, leveraged by the expected widespread availability of GPS-capable wireless devices in the near future. We describe procedures for approximating irregular handoff-request boundary, and for estimating a mobile's remaining time to handoff and target handoff cell. Guard bandwidth is adjusted on the fly only when a handoff is anticipated to occur shortly. The scheme is robust against modifications in handoff parameters, as well as changes in terrain and manmade features that could affect radio propagation. It also caters for heterogeneous bandwidth requirements. Admission control algorithms for both new calls and handoff calls are provided. Simulation results show that our scheme is able to meet specified forced termination probability under high offered load, and that resources are not reserved unnecessarily.

Amidst the rapid growth of wireless broadband networks, little attention has been paid to wireless network survivability issues. We propose a survivable wireless ATM network architecture that allows it to survive from a single base station failure condition, by redirecting a failure cell's traffic via its six neighboring cells. We present two failure-handling schemes for the proposed architecture. The first scheme is a bandwidth reservation (BR) scheme that is targeted at achieving high survivability by reserving bandwidths at appropriate locations in the network. The second scheme is a best-effort (BE) scheme that does not perform any reservation, and is targeted at achieving high bandwidth utilization. Simulation results show that the BR scheme achieves good survivability as expected. On the other hand, the BE scheme provides better utilization while having slightly lower average survivability. The decrease in average survivability for the BE scheme is not tremendous, largely due to its flexibility in the use of spare bandwidths from neighboring cells when a failure occurs. However, the BE scheme requires more frequent update messages between the switches in order to update each other about the amount of bandwidth that they could provide for failure-handling. We have also considered important issues, such as switchover time and data integrity, for our proposed schemes.

This paper addresses an optimal link capacity design problem for self-healing asynchronous transfer mode (ATM) networks based on two different restoration schemes: line restoration and end-to-end restoration. Given a projected traffic demand, capacity and flow assignment is jointly optimized to find an optimal capacity placement. The problem can be formulated as a large-scale linear programming. The basis matrix can be readily factorized into an LU form by taking advantage of its special structure, which results in a substantial reduction on the computation time of the revised simplex method. A row generation and deletion mechanism is developed to cope with the explosive number of constraints for the end-to-end restoration-based networks. In self-healing networks, end-to-end restoration schemes have been considered more advantageous than line restoration schemes because of a possible reduction of the redundant capacity to construct a fully restorable network. A comparative analysis is presented to clarify the benefit of end-to-end restoration schemes quantitatively in terms of the minimum resource installation cost. Several networks with diverse topological characteristics as well as multiple projected traffic demand patterns are employed in the experiments to see the effect of various network parameters. The results indicate that the network topology has a significant impact on the required resource installation cost for each restoration scheme. Contrary to a wide belief in the economic advantage of the end-to-end restoration scheme, this study reveals that the attainable gain could be marginal for a well-connected and/or unbalanced network.

In self-healing networks, end-to-end restoration schemes have been considered more advantageous than line restoration schemes because of a possible cost reduction of the total capacity to construct a fully restorable network. This paper clarifies the benefit of end-to-end restoration schemes quantitatively through a comparative analysis of the minimum link capacity installation cost. A jointly optimal capacity and flow assignment algorithm is developed for the self-healing ATM networks based on end-to-end and line restoration. Several networks with diverse topological characteristics as well as multiple projected traffic demand patterns are employed in the experiments to see the effect of various network parameters. The results indicate that the network topology has a significant impact on the required resource installation cost for each restoration scheme. Contrary to a wide belief in the economic advantage of the end-to-end restoration scheme, this study reveals that the attainable gain could be marginal for a well-connected and/or unbalanced network.

This paper describes an approach for the analysis of multistage switching networks with a variety of buffer sharing strategies. The approach allows non-uniform bursty traffic, and it features a computational complexity which is independent of the buffer size. We decompose the complex shared buffer analysis problem into an equivalent dedicated buffer problem through an iterative series of buffer size approximations. Results are compared with simulations and are used to quantify the performance differences of several buffer sharing policies.

The advent of high-capacity optical fiber has increased the impact of a network failure in high-speed networks since a large volume of data can be lost even in a short outage. Self-healing algorithms have previosly been proposed to achieve fast restoration from a failure, but their success greatly depends on how traffic is distributed and how spare capacity is dimensioned over the network when a failure happens. Thus, in order to offer better network survivability, it is crucial that a network manager realizes a restorable traffic assignment in response to changing traffic demand and facility network configuration. The authors address the problem of virtual path routing for survivable asynchronous transfer mode (ATM) networks. An algorithm is developed to find a virtual path configuration and bandwidth assignment that minimizes the expected amount of lost flow upon restoration from a network failure. The concept of two-step restoration is introduced to achieve fast restoration as well as optimal reconfiguration. The problem can be formulated as a nonlinear, nonsmooth multicommodity flow problem with linear constraints. A modified flow deviation method is developed to obtain a near-optimal solution, where premature convergence to a nonsmooth point could be avoided by adjusting an optimization parameter. The result of the performance evaluation indicates that the proposed routing scheme can detect the links that are vulnerable to a failure under the current traffic demand pattern and adjust a flow so as to improve the network survivability level.

The Tera ATM LAN project at Carnegie Mellon University addresses the interconnection of hundreds of workstations in the Electrical and Computer Engineering Department via an ATM-based network. The Tera network architecture consists of switched Ethernet clusters that are interconnected using an ATM network. This paper presents the Tera network architecture, including an Ethernet/ATM network interface, the Tera ATM switch, and its performance analysis. The Tera switch architecture for asynchronous transfer mode (ATM) local area networks (LAN's) incorporates a scalable nonblocking switching element with hybrid queueing discipline. The hybrid queueing strategy includes a global first-in first-out (FIFO) queue that is shared by all switch inputs and dedicated output queues with small speedup. Due to hybrid queueing, switch performance is comparable to output queueing switches. The shared input queue design is scalable since it is based on a Banyan network and N FIFO memories. The Tera switch incorporates an optimal throughput multicast stage that is also based on a Banyan network. Switch performance is evaluated using queueing analysis and simulation under various traffic patterns.>

Addresses the problem of virtual path routing for survivable ATM networks. An algorithm is developed to find a virtual path configuration and bandwidth assignment which minimizes the expected amount of lost flow upon network restoration from a failure. The concept of two-step restoration is introduced to achieve fast restoration as well as optimal reconfiguration. The problem can be formulated as a nonlinear nonsmooth multicommodity flow problem with linear constraints. A modified flow deviation approach is developed to obtain the steepest decent direction for a nonsmooth objective function. Based on the proposed scheme, convergence to a near-optimum is possible by properly adjusting optimization parameters.>

A new ATM switch architecture is presented. Our proposed Multinet switch is a self-routing multistage switch with partially shared internal buffers capable of achieving 100% throughput under uniform traffic. Although it provides incoming ATM cells with multiple paths, the cell sequence is maintained throughout the switch fabric thus eliminating the out-of-order cell sequence problem. Cells contending for the same output addresses are buffered internally according to a partially shared queueing discipline. In a partially shared queueing scheme, buffers are partially shared to accommodate bursty traffic and to limit the performance degradation that may occur in a completely shared system where a small number of calls may hog the entire buffer space unfairly. Although the hardware complexity in terms of number of crosspoints is similar to that of input queueing switches, the Multinet switch has throughput and delay performance similar to output queueing switches.>

An asynchronous transfer mode (ATM) switch architecture, the multinet switch, is presented. It is a self-routing multistage switch with internal buffers capable of achieving 100% throughput. Although it provides incoming ATM cells with multiple paths, the cell sequence is maintained throughout the switch fabric, thus eliminating the out-of-order cell sequence problem. Cells contending for the same output addresses are buffered internally according to a partial buffer sharing discipline. In a partial buffer sharing scheme, buffers are partially to accommodate bursty traffic and to limit the performance degradation that may happen in a completely shared system where a small number of calls ties up the entire buffer space unfairly. Although the hardware complexity is similar to that of the input queueing switches, the multinet switch has throughput and delay performance similar to that of output queueing switches without the hardware complexity. A simple extension of the multinet switch to handle multiple priority traffic and multicast traffic is proposed.>

A new nonblocking copy network is presented, for use in an ATM switch supporting BISDN, with a shared-memory input buffer. Blocked cells from any switch input are stored in a single shared input buffer. The copy network consists of three Omega networks and shared-memory queues. The design is scalable for large numbers of inputs due to a low hardware complexity, O(N log/sub 2/ N), and distributed operation and control. It is shown by simulation results that a switch incorporating the shared-memory copy network has increased throughput and lower buffer requirements to maintain low packet loss probability when compared to a switch with a discrete buffer copy network.>

A simple, highly flexible admission control algorithm for ATM-based B-ISDN, known as the effective-bit-rate/table-lookup (EBR/TL) method, is proposed. An EBR method works by allocating bandwidth using a percentage, a(0>

The authors present a new nonblocking property of the reverse banyan network under a particular input packet pattern at the input ports. The reverse banyan network is the mirror image of the banyan network. If the input packets of the N*N reverse banyan network have consecutive output address as modulo N, then the reverse banyan network is nonblocking. The routing of packets in the reverse banyan network is described, and the nonblocking property of the reverse banyan network is proved. A possible application of this property in the switching network is discussed.>

A report is presented on a queueing analysis and a simulation study of a switch fabric based on a buffered banyan structure whereby buffers are placed at the output links of each switching element. When buffers are located at the input links, it is well known that maximum throughput is limited to approximately 0.45 under a uniform input traffic pattern. This bottleneck is due to the head of the line (HOL) contention at each switching element and is intrinsic to input queueing. The authors propose a buffered banyan switch built from smaller knockout switches which are output-buffered switches. With small knockout switches as the basic switching elements, the complexity of the overall switch fabric is manageable and no internal clock speedup is required. Furthermore, it is shown that with the proposed output-buffered banyan switch, a maximum throughput of 1 can be achieved.>

An asynchronous transfer model (ATM) switch which has lower hardware complexity than that of the output queuing switch is presented. The reduction in the hardware complexity is obtained without either losing the self-routing property or suffering from performance degradation under nonuniform traffic patterns. The switch consists of shift networks that are interconnected with distributors in two stages. Introducing the distributor between the stages of shift networks allows of packets to be distributed evenly to all the input ports of shift networks in the following stage. Although the switch becomes blocking, it retains the self-routing property and achieves the maximum throughput of 100% with only a small additional delay.>

A switching network that approaches a maximum throughput of 100% as buffering is increased is proposed. This self-routing switching network consists of simple 2*2 switching elements, distributors, and buffers located between stages and in the output ports. The proposed switching requires a speedup factor of two. The structure and the operation of the switching network are described, and its performance is analyzed. The switch has log/sub 2/N stages that move packets in a store-and-forward fashion, incurring a latency of log/sub 2/N time periods. The performance analysis of the switch under uniform traffic pattern shows that the additional delay is small, and a maximum throughput of 100% is achieved as buffering is increased.>

An analytical method of evaluating the performance of the buffered banyan packet-switching network under nonuniform traffic patterns is presented. It is shown that nonuniform traffic can have a detrimental effect on the performance of the network. The analytical model is extended to evaluate the performance of multibuffer and parallel banyan networks. These modified networks are shown to have better throughput capacity and delay performance than the single-buffer banyan network.>

The authors present an analytical method to evaluate the performance of the buffered Banyan packet-switching network under nonuniform traffic patterns. It is shown that the nonuniform traffic can have a detrimental effect on the performance of the network. The analytical model is extended to evaluate the performance of multibuffer and parallel Banyan networks. These modified networks are shown to have better throughput capacity and delay performance than the single-buffer Banyan network.>