Search

We present a mathematical model for a distributed process of routing optimization that could be run in the control plane of the Internet using existing EGP routing protocols. A more detailed description of the results presented in this paper is given in [1]. [ABSTRACT FROM AUTHOR]

The challenges of significant network dynamics and limited bandwidth capacities have to be considered when designing efficient algorithms for distributing large volumes of content in overlay networks. This paper presents Outburst, a novel approach for overlay content distribution based on rateless codes. In Outburst, we code content bitstreams with rateless codes at the source, and take advantage of the superior properties of rateless codes to provide resilience against network dynamics and node failures. We recode the bitstreams at each receiver node, so that the need for content reconciliation in parallel downloading is eliminated, and the delivery of redundant content is minimized. The effectiveness and efficiency of Outburst are demonstrated with simulations. [ABSTRACT FROM AUTHOR]

IETF working groups developed mobile IP protocols to support host mobility, and the NEMO(Network Mobility) Working Group, specifically, developed the NEMO Basic Support Protocol. The NEMO basic solution enables mobile networks to change their point of attachment to the Internet. The protocol, however, leads to suboptimal route, and other problems. The problems become serious when mobile networks are nested. In this paper, we present a route optimization mechanism for nested mobile networks based on ‘Limited Prefix Delegation' technique. We present performance evaluation results by simulation to confirm the effectiveness. [ABSTRACT FROM AUTHOR]

In multi access network environments, mobile stations may encounter multiple choices for selecting an access network. Carefully designed access selection schemes can provide not only mobile users with better services but also network operators with better resource utilizations. It is also envisioned that further improvements can be achieved by redistributing mobile stations from one access network to anther (i.e., vertical handovers). Such decisions should be made by following carefully designed, yet simple to implement, protocols. In this paper, we present a cost-based scheme for access selection and vertical handover decision algorithms. The proposed algorithm was implemented in a Java-based simulator called MANSim (Multiple Access Network Simulator) that we developed. Our simulation results show that during the congested periods, the network throughput is significantly improved with greatly reduced call drop rate. [ABSTRACT FROM AUTHOR]

This paper proposes a network memory-based P2P IO BUffering Service (PIBUS), which buffers blocks for IO-intensive applications in P2P network memory like a 2-level disk cache. PIBUS reduces the IO overhead when local cache is missed due to speed advantage of network memory over disks, and improves hit ratio based on accurate classification of IO behaviors. [ABSTRACT FROM AUTHOR]

There is a growing interest in virtualized network infrastructures as a means to enable experimental evaluation of new network architectures on a realistic scale. The National Science Foundation's GENI initiative seeks to develop a national experimental facility that would include virtualized network platforms that can support many concurrent experimental networks, with the goal of reducing barriers to new network architectures. This paper focuses on how to extend the concept of virtualized networking through LAN-based access networks to the end systems. We demonstrate that our approach can improve performance by an order of magnitude over other approaches and can enable virtual networks that provide end-to-end quality of service. [ABSTRACT FROM AUTHOR]

Trying to find a practical solution for QoS provision capabilities over IP networks has been subject of a great research effort during last years. The main aim of the proposal presented in this paper is to allow QoS provision over IP networks without the requirement of upgrading every router of the Internet. Upgrading some routers within the network is still needed, but no constraints do exist related to which routers might be upgraded. [ABSTRACT FROM AUTHOR]

For networks providing QoS guarantees, this paper determines and evaluates the worst case end-to-end delays for strategies based on network coding and multiplexing. It is shown that the end-to-end delay does not depend on the same parameters with the two strategies. This result can be explained by the fact that network coding can cope with congestions better than classical routing because it processes simultaneously packet from different flows. In counterpart, additional delays like algebraic combinations of packets are added. [ABSTRACT FROM AUTHOR]

Ethernet passive optical network (EPON) has been considered to solve the last mile bottleneck problem. In an effort to accommodate the explosive bandwidth demands from subscribers, the optical line terminal (OLT) efficiently divides and allocates time slots for data upstream to all optical network units (ONUs) in EPON. This technology is expected to be a core in the future fiber-to-the-home/-office/-curb (FTTH/O/C). We study previous algorithms for dynamic bandwidth allocation (DBA) in interleaved polling with adaptive cycle time (IPACT). For effective bandwidth allocation of the uplink channel, we propose an adaptive window-tuning algorithm (AdWin) based on the excessive bandwidth. This algorithm not only satisfies bandwidth demands of ONUs within the possible scope, but also seeks fairness among ONUs. The comprehensive computer simulation results indicate that the proposed scheme is up to 94% and 94% lower than previous schemes in terms of average packet delay and average queue size, respectively. It also demonstrates up to 86% improved performance in regards to packet loss ratio. [ABSTRACT FROM AUTHOR]

To address the burst loss problem in OBS network, we propose a new optical burst control algorithm, which estimates the future burst traffic condition to eliminate the effect of congestion reaction delay. Through the simulations, we verify that the proposed algorithm outperforms other existing burst control algorithms. [ABSTRACT FROM AUTHOR]

Internet instability, also referred to as route flaps, can propagate to the whole Internet and consume remarkable computational resource of the routers. Route Flap Damping (RFD) [1] is designed to stabilize the Internet by suppressing persistent route flaps. RFD is a penalty basedmechanism.Themagnitude of the penalty value indicates the degree of instability of an inter-domain route.Once the penalty reaches a certain threshold, the route will be suppressed. This simple mechanism does not work perfectly. First, the way it identifies route flaps or accumulates penalty is too aggressive and may suppress a fairly stable route with a few occasional flaps. Second, a route may be suppressed even after it has converged. [ABSTRACT FROM AUTHOR]

This paper focuses on the design of IBGP networks, which are very important to the reliability and stability of Internet. Although several metrics have been presented to measure the robustness of IBGP networks, they only considered the impact of route reflection networks on the control plane. A robust network should have low sensitivity to traffic load variations. So we propose a new metric to characterize the impact of route reflection networks on the data plane, which is called TDR (Traffic Diversion Rate). Simulation results show that adopting the optimal route reflection topology that minimizes TDR will make the network lose or shift much less traffic, compared with adopting the optimal route reflection topologies found according to other metrics. [ABSTRACT FROM AUTHOR]

In this paper, we analyze traffic seen at public WLANs "in the wild" where we do not have access to any of the backend infrastructure. We study six such traces collected around Portland, Oregon and conduct an analysis of fine time scale (second or fraction of a second) packet, flow, and error characteristics of these networks. [ABSTRACT FROM AUTHOR]

We assess the performance of a partially shared bottleneck buffer for scalable video. The arrival process of the video packets is modelled by means of a two-class discrete batch Markovian arrival process. Using a matrix-analytic approach, we retrieve various performance measures. We illustrate our approach by means of a numerical example. [ABSTRACT FROM AUTHOR]

Followed by the popularity of the Internet, a number of access technologies to the Internet have been developed. EAP is an authentication framework. It is designed to provide the authentication functionality in the access network. Because of its flexibility and extensibility EAP poses a global solution for the authentication supported by many access networks. However, EAP has critical weaknesses in the protocol which may, in turn, decrease the EAP performance. Some of the weaknesses are caused by the "lock-step" flow control which only supports a single packet in flight. Considering the weaknesses, we propose the solution for the flow control. Using simulation we prove that our solutions improve the EAP performance. [ABSTRACT FROM AUTHOR]

Network monitoring is essential to the correct and efficient operation of overlay networks, and active measurement is a key design problem in network monitoring. Unfortunately, almost all active probing algorithms ignore the measurement conflict problem: Active measurements conflict with each other - due to the nature of these measurements, the associated overhead, and the network topology - which results in reporting incorrect results. In this paper, we consider the problem of scheduling periodic QoS measurement tasks in overlay networks. We first show that this problem is NP-complete, and then propose a conflict-aware scheduling algorithm whose goal is to maximize the number of measurement tasks that can run concurrently, based on a well known approximation algorithm. Simulation results show that our algorithm achieves 25% better schedulability over the existing algorithm. Finally, we discuss various practical considerations, and identify several interesting research problems in this context. [ABSTRACT FROM AUTHOR]

In this paper, we present a measurement based characterization of the Ultra Wideband (UWB) channel in a data center environment. We find that although a modified Saleh-Valenzuela model characterizes the UWB channel, some of the model parameters such as delay spread and log normal shadowing are unique to the data center environment. [ABSTRACT FROM AUTHOR]

The IEEE 802.15.4 standard provides the widely accepted solution for low-cost and low-power wireless communications. Despite its design support for low duty-cycle operation, the fixed superframe size in Beacon-enabled mode limits its capabilities due to two contrasting goals; energy efficiency and higher data throughput. In this paper, we propose an enhancement of IEEE 802.15.4 Beacon-enabled mode which adaptively adjusts the active period based on the traffic information. In order to detect the data traffic in the networks, the proposed scheme utilizes the IEEE 802.15.4 CCA function. Evaluation results show that our scheme can improve energy efficiency as well as data throughput. [ABSTRACT FROM AUTHOR]

While many existing rate adaptation schemes in IEEE 802.11 Wireless LANs result in severe throughput degradation since they do not consider the collision effect when selecting the transmission rate, CARA (Collision-Aware Rate Adaptation) [1] shows improved system performance thanks to its collision-awareness capability. In this paper, we propose two enhancements to the original CARA scheme to further improve the system performance. The first one is called CARA-RI, which extends CARA's collision-awareness capability in making rate increase decisions, while the second one, called CARA-HD, incorporates a hidden station detection mechanism. Simulation results show that the proposed schemes outperform the original CARA significantly under various randomly-generated network topologies. [ABSTRACT FROM AUTHOR]

Building robust network monitoring applications is hard given the unpredictable nature of network traffic. Complex analysis on streaming network data usually leads to overload situations when presented with anomalous traffic, extreme traffic mixes or highly variable rates. We present an on-line predictive load shedding scheme for monitoring systems that quickly reacts to overload situations by gracefully degrading the accuracy of analysis methods. The main novelty of our approach is that it does not require any knowledge of the monitoring applications. This way we preserve a high degree of flexibility, increasing the potential uses of these systems. We implemented our scheme in an existing network monitoring system and deployed it in a research ISP network. Our experiments show a 10-fold improvement in the accuracy of the results during long-lived executions with several concurrent monitoring applications. The system efficiently handles extreme load situations, while being always responsive and without undesired packet losses. [ABSTRACT FROM AUTHOR]

Proportional Fair (PF), a frequently used scheduling algorithm in 3G wireless networks, can unnecessarily starve "well-behaved" users in practice. One of the main causes behind PF-induced starvation is its inability to distinguish between users who are backlogged and users who are not. In this paper, we describe how a simple parallel PF instance can mitigate such starvation. [ABSTRACT FROM AUTHOR]

Two distributed algorithms DMMT-OA and DMMT-DA have been recently proposed to maximize the multicast lifetime for directional communications in wireless ad-hoc networks. The experimental results have shown their superior performance than other centralized algorithms; however, their theoretical performance in terms of approximation ratio is still unknown. In this paper, we use graph theoretic approach to derive the approximation ratio for both algorithms. Furthermore, we have discovered by the first time that both ratios are bounded by a constant number. [ABSTRACT FROM AUTHOR]

To exploit multi-rate capability as well as improve performance in wireless local area networks (WLANs), many mechanisms were proposed on IEEE 802.11 media access control (MAC) layer. However, no effort has been invested to exploit the multi-rate capability for power saving mechanism in MAC layer. In this paper, we propose a Voluntary Relaying MAC Protocol, called VRMP, to achieve both performance improvement and power saving by leveraging the multi-rate capability. In voluntary relaying scheme, if a node can support low rate node's data packet at higher rate and has sufficient power, after cooperatively sending data packet at higher rate, all nodes go into sleep mode as quickly as possible to reduce power consumption. Simulation results show that the VRMP improves throughput by 30 ~ 60% as well as reduces power consumption by 10 ~ 55% than the legacy mechanism. [ABSTRACT FROM AUTHOR]

The impact of capture effect on the IEEE 802.11 networks with different transmission speeds is investigated in this paper. A new Markov chain model considering capture effect for the binary exponential back-off scheme in the MAC layer has been proposed in the first time. Based on the new Markov chain model, a new throughput model is proposed, and then the impact of capture effect on throughput has been analyzed in the condition of different transmission speeds. The performance analysis shows that, in the RTS/CTS scheme, the improvement throughput of high speed networks caused by capture effect is more than that of low speed networks. [ABSTRACT FROM AUTHOR]

The justification for using mutihop clusters may be found in [1]. In the well known heuristic proposed in [2], the d-dominating set of clusterheads is first selected by using nodes identifiers and then clusters are formed. In this paper we generalise this algorithm in order to select nodes depending of a given criterion (as the the degree, density or energy of nodes). The first section of this paper simplifies and proves the correctness of our generalised algorithm to select clusterheads. The cluster formation process proposed in [2] is extensively studied in the second section and is proved to be false. [ABSTRACT FROM AUTHOR]

For an ad hoc IEEE 802.11 WLAN we investigate how the stations' incentives to launch a backoff attack i.e., to configure small minimum and maximum CSMA/CA contention windows in pursuit of a larger-than-fair bandwidth share, affect a proposed capacity-fairness index (CFI). We link CFI to the network size, "power awareness," a station's perception of the other stations' susceptibility to incentives, and the way of learning how the other stations perceive the other stations' susceptibility to incentives. [ABSTRACT FROM AUTHOR]

Traditionally, route discovery in MANETs operates in unilateral (source-initiated) manner. We propose a new scheme called bilateral route discovery (BRD), where both source and destination actively participate in a route discovery process. BRD has the potential to reduce the control overhead by one half. As an underlying protocol for BRD, we propose gratuitous route error reporting (GRER) to notify the destination of a broken route. The destination can thus play an active role in the upcoming route re-discovery. [ABSTRACT FROM AUTHOR]

Mobile ad-hoc networks (MANETs) utilize broadcast channels, where wireless transmissions occur from one user to many others. In a broadcast channel the same transmission can lead to different information rates to different users depending on the channel capacity between the transmitter and receiver pair. According to coding theory, there is a certain channel capacity that limits the rate of information that can be sent through the channel. Thus, different channel capacities result in different acceptable rates for the users. In this paper, we utilize a superposed coding scheme in a MANET scenario to provide different rates for users with different channel capacities using a single broadcast transmission. We have created techniques to extend this multi-rate concept to network-wide broadcasting scenarios, providing the ability for nodes to appropriately trade-off delay vs. quality. We describe our approach and provide simulation results showing the benefits and limitations of superposed coding in network-wide broadcasting. [ABSTRACT FROM AUTHOR]

This paper describes SEA-LABS (Sensor Exploration Apparatus utilizing Low-power Aquatic Broadcasting System), a low-cost, power-efficient Wireless Sensor Network (WSN) for sustained, real-time monitoring of shallow water coral reefs. The system is designed to operate in remote, hard-to-access areas of the world, which limits the ability to perform on-site data retrieval and periodic system maintenance (e.g., battery replacement/recharging). SEA-LABS thus provides a customized solution to shallow-water environmental monitoring addressing the trade-offs between power conservation and the system's functional requirements, namely data sensing and processing as well as real-time, wireless communication. We present SEA-LABS' architecture and its current implementation. Finally, we share our experience deploying SEA-LABS in the Monterey Bay. [ABSTRACT FROM AUTHOR]

In an effort to prosecute P2P users, RIAA and MPAA have reportedly started to create decoy users: they participate in P2P networks in order to identify illegal sharing of content. This has reportedly scared some users who are afraid of being caught. The question we attempt to answer is how prevalent is this phenomenon: how likely is it that a user will run into such a "fake user" and thus run the risk of a lawsuit? The first challenge is identifying these "fake users". We collect this information from a number of free open source software projects which are trying to identify such IP address ranges by forming the so-called blocklists. The second challenge is running a large scale experiment in order to obtain reliable and diverse statistics. Using Planetlab, we conduct active measurements, spanning a period of 90 days, from January to March 2006, spread over 3 continents. Analyzing over a 100 GB of TCP header data, we quantify the probability of a P2P user of being contacted by such entities. We observe that 100% of our nodes run into entities in these lists. In fact, 12 to 17% of all distinct IPs contacted by any node were listed on blocklists. Interestingly, a little caution can have significant effect: the top five most prevalent blocklisted IP ranges contribute to nearly 94% of all blocklisted IPs and avoiding these can reduce the probability of encountering blocklisted IPs to about 1%. In addition, we examine other factors that affect the probability of encountering blocklisted IPs, such as the geographical location of the users. Finally, we find another surprising result: less than 0.5% of all unique blocklisted IPs contacted are owned explicitly by media companies. [ABSTRACT FROM AUTHOR]

In recent years, massively multiplayer online games (MMOGs) have become very popular by providing more entertainment and sociability than single-player games. In order to prevent cheaters to gain unfair advantages in peer-to-peer (P2P)-based MMOGs, several cheat-proof schemes have been proposed by using digital signatures. However, digital signatures generally require large amount of computations and thus may not be practical to achieve real-time playability. We propose an Efficient and Secured Event Signature (EASES) protocol to efficiently sign discrete event messages. Most messages need only two hash operations to achieve non-repudiation and event agreement. The computation, memory, and bandwidth consumptions of EASES are low, which makes it applicable to P2P-based MMOGs. [ABSTRACT FROM AUTHOR]

With DoS/DDoS attacks emerging as one of the primary security threats in today's Internet, the search is on for an efficient DDoS defense mechanism that would provide attack prevention, mitigation and traceback features, in as few packets as possible and with no collateral damage. Although several techniques have been proposed to tackle this growing menace, there exists no effective solution to date, due to the growing sophistication of the attacks and also the increasingly complex Internet architecture. In this paper, we propose an unified framework that integrates traceback and mitigation capabilities for an effective attack defense. Some significant aspects of our approach include: (1) a novel data cube model to represent the traceback information, and its slicing along the lines of path signatures rather than router signatures, (2) characterizing traceback as a transmission scheduling problem on the data cube representation, and achieving scheduling optimality using a novel metric called utility, (3) and finally an information delivery architecture employing both packet marking and data logging in a distributed manner to achieve faster response times. The proposed scheme can thus provide both per-packet mitigation and multi-packet traceback capabilities due to effective data slicing of the cube, and can attain higher detection speeds due to novel utility rate analysis. We also contrast this unified scheme with other well-known schemes in literature to understand the performance tradeoffs, while providing an experimental evaluation of the proposed scheme on real data sets. [ABSTRACT FROM AUTHOR]

Active measurements are a useful tool for obtaining a variety of Internet metrics. One-way metrics, in general, require the execution of processes at the remote machine and/or machines with synchronized clocks. This work proposes a new algorithm to estimate the first two moments of the one-way delay random variable without the need to access a target machine and to have the machine clocks synchronized. The technique uses the IPID field information and can be easily implemented using ICMP Echo request and reply messages. [ABSTRACT FROM AUTHOR]

We re-examine the two reasons for the conservative 1-second Minimum TCP-RTO to protect against spurious timeouts: i) the OS clock granularity and ii) the Delayed ACKs. We find that reason (i) is canceled in modern OSs; we carefully design a mechanism to deal with reason (ii). Simulation results show that in next generation's high-speed, wireless-access networks, TCP-RTO should not be limited by a fixed, conservative lower bound. [ABSTRACT FROM AUTHOR]

In this paper, we propose a practical and efficient technique, Forecaster, to estimate (1) the end-to-end available bandwidth, and (2) the speed of the most congested (tight) link along an Internet path. Forecaster is practical since it does not assume any a priori knowledge about the measured path, does not make any simplifying assumptions about the nature of cross-traffic, does not assume the ability to capture accurate packet dispersions or packet queueing delays, and does not try to preserve inter-packet spacing along path segments. It merely relies on a simple binary test to estimate whether each probe packet has queued in the network or not. Forecaster is efficient as it only requires two streams of probe packets that are sent end-to-end at rates that are much lower than the available bandwidth of the investigated path, thus avoiding path saturation. Theoretical analysis and experimental results validate the efficacy of the proposed technique. [ABSTRACT FROM AUTHOR]

Most Internet-based collaborative computing systems face the major problem: freeriding. The abundance of freeriders, and load imbalance it creates, punishes those peers who do actively contribute to the network by forcing them to overuse their resources. Hence, the overall system performance becomes to degrade quickly. The goal of this paper is aimed to provide an efficient approach to distinguish the dishonest peers from the honest peers. The key idea of our approach is to make use of the relationship between the perceived throughput and the available bandwidth. First, we do a comprehensive study of available bandwidth estimation tools. Next, we propose integrity-aware bandwidth guarding algorithm, which is designed according to the perceived throughput and the available bandwidth estimation. Finally, the simulation results illustrate that our approach can correctly identify dishonest peers and be of great help in constructing a better overlay structure for many peer-to-peer and multicast applications. [ABSTRACT FROM AUTHOR]

For many network services, such as firewalling, load balancing, or cryptographic acceleration, data packets need to be classified (or filtered) before network appliances can apply any action processing on them. Typical actions are header manipulations, discarding packets, or tagging packets with additional information required for later processing. Structured data, such as XML, is independent from any particular presentation format and is an ideal information exchange format for a variety of heterogeneous sources. In this paper, we propose a new algorithm for fast and efficient classification of structured data in the network. In our approach, packet processing and classification is performed on structured payload data rather than only packet header information. Using a combination of hash functions, Bloom filter, and set intersection theory our algorithm builds a hierarchical and layered data element tree over the input grammar that requires logarithmic time and tractable space complexity. [ABSTRACT FROM AUTHOR]

Our Token Based Switch (TBS) implementation shows that a packet-based admission control system can be used to dynamically select a fast end-to-end connection over a hybrid network at gigabit speeds. TBS helps high-performance computing and grid applications that require high bandwidth links between grid nodes to bypass the regular Internet for authorised packets by establishing shortcuts on network links with policy constraints. TBS is fast and safe and uses the latest network processor generation (Intel IXP2850) and the Fairly Fast Packet Filter software framework. [ABSTRACT FROM AUTHOR]

TCP is shown to be inefficient and instable in high speed and long latency networks. The eXplicit Control Protocol (XCP) is a new and promising protocol that outperforms TCP in terms of efficiency, stability, queue size, and convergence speed. However, Low et al. recently discovered a weakness of XCP. In a multi-bottleneck environment, XCP may achieve as low as 80% utilization at a bottleneck link and consequently some flows may only receive a small fraction of their max-min fair rates. This paper proposes iXCP, an improved version of XCP. Extensive simulations show that iXCP overcomes the weakness of XCP, and achieves efficient and fair bandwidth utilization in both single- and multi- bottleneck environments. In addition, we prove that iXCP is max-min fair in steady state. This result implies that iXCP is able to fully utilize bottleneck bandwidth. Simulations also show that iXCP preserves the good properties of XCP, including negligible queue lengths, near-zero packet loss rates, scalability, and fast convergence. [ABSTRACT FROM AUTHOR]

We study online multicasting in WDM networks with shared light splitter bank. Our objective is to maximize the network throughput. It is desirable that the cost of realizing each multicast request be minimized, and the network throughput will be maximized ultimately through the cost saving on each individual request. We first propose a cost model for realizing an online multicast request under such network environments with limited light splitters and wavelength converters, which models the cost of utilization of network resources, particularly in modelling the light splitting and wavelength conversion ability at nodes. We then show that finding a cost-optimal multicast tree for a multicast request under the proposed cost model is NP-complete, and instead devise approximation and heuristic algorithms for it. We finally conduct experiments to evaluate the performance of the proposed algorithms. The results show that the proposed algorithms are efficient and effective in terms of network throughput. [ABSTRACT FROM AUTHOR]

The majority of Internet users rely on the Transmission Control Protocol (TCP) to download large multimedia files from remote servers (e.g. P2P file sharing). TCP has been advertised as a fair-share protocol. However, when session round-trip-times (RTTs) radically differ from each other, the share (of the bottleneck link) may be anything but fair. This motivates us to explore a new TCP, TCP Libra, that guarantees fair sharing regardless of RTT. The key element of TCP Libra is the unique window adjustment algorithm that provably leads to RTT-independent throughput, yet converging to the fair share. We position TCP Libra in a non-linear optimization framework, proving that it provides fairness (in the sense of minimum potential delay fairness) among TCP flows that share the same bottleneck link. Equally important are the friendliness of Libra towards legacy TCP and the throughput efficiency. TCP Libra is source only based and thus easy to deploy. Via analytic modeling and simulations we show that TCP Libra achieves fairness while maintaining efficiency and friendliness to TCP New Reno. A comparison with other TCP versions that have been reported as RTT-fair in the literature is also carried out. [ABSTRACT FROM AUTHOR]

Intelligent Route Control (IRC) technologies allow multihomed networks to dynamically select egress links based on performance measurements. TCP congestion control, on the other hand, dynamically adjusts the send-window of a connection based on the current path's available bandwidth. Little is known about the complex interactions between IRC and TCP congestion control. In this paper, we consider a simple dual-feedback model in which both controllers react to packet losses, either by switching to a better path (IRC) or by reducing the offered load (TCP congestion control). We first explain that the IRC-TCP interactions can be synergistic as long as IRC operates on larger timescales than TCP ("separation of timescales"). We then examine the impact of sudden RTT changes on TCP, the behavior of congestion control upon path changes, the effect of IRC measurement delays, and the conditions under which IRC is beneficial under two path impairment models: short-term outages and random packet losses. [ABSTRACT FROM AUTHOR]

We propose a scheme for providing differentiated services in an optical burst switching network environment. Our framework considers differentiation among several traffic classes in packet loss probability, assuming only that the network differentiates neatly between two classes of bursts. We formulate a reduced load fixed point model for evaluating the blocking probabilities of the different types of bursts. The model is the basis to compare several flavors of OBS in terms of ability for achieving quality of service. Motivated by technological constraints, we investigate the effect of wavelength density on its performance. [ABSTRACT FROM AUTHOR]

This paper presents LiTGen, an easy to use and tune open-loop traffic generator that statistically models IP traffic on a per user and application basis. From a packet level capture originating in an ISP wireless network, and taking the example of Web traffic, we show that the simple underlying structure of LiTGen is sufficient to reproduce the traffic burstiness accurately. In addition, the flexibility of LiTGen enables us to investigate the sensitivity of the traffic structure with respect to the distributions of the random variables involved in the model, and their possible dependencies. [ABSTRACT FROM AUTHOR]

Overlay topology design has been one of the most challenging research areas over the past few years. In this paper, we consider the problem of finding the overlay topology that minimizes a cost function which takes into account the overlay link creation cost and the routing cost. First, we formulate the problem as an Integer Linear Programming (ILP) given a traffic matrix in case of cooperative and non cooperative node behavior. Then, we propose some heuristics to find near-optimal overlay topologies with a reduced complexity. The solutions of the ILP problem in average-size networks have been analyzed, showing that the traffic demands between the nodes affects the decision of creating new overlay links. The heuristics are also compared through extensive numerical evaluation, and guidelines for the selection of the best heuristic as a function of the cost parameters are also provided. [ABSTRACT FROM AUTHOR]

This paper investigates a new technique called Bayesian-Block-Analysis (BBA) for analyzing the time varying rate of events. The first goal is to evaluate the accuracy of BBA in identifying the rate changes in synthetic traces that have a given interevent times distribution, and known rate change points. We find that BBA is highly accurate on traces with exponential interevent times and known rate changes, and reasonably accurate with more heavier-tailed interevent times. The second goal is to apply BBA to actual network event traces. And for request arrivals or loss rate traces, BBA identifies significant stationary-rate periods which are qualitatively consistent with previous results obtained with less efficient or less accurate techniques. For packet arrivals to gateways, BBA identifies stationary rate periods that are corroborated by binning the data on a new timescale. Finally, we also show BB-online rate estimation is accurate for synthetic as well as actual system traces. [ABSTRACT FROM AUTHOR]

This paper discusses the influence of the maturity of the photonic component industry (PCI) in the business prospects of optical access network deployments. Using the TONIC techno-economic tool, the business prospects of such deployments (in terms of traditional investment indexes such as the Net Present Value - NPV or Internal Rate of Return - IRR) are related to the several factors which characterize the maturity of the PCI such as production volumes and component cost evolution. The analysis shows that even if the cost of Fiber-to-the-Home/Office (FTTH/O) and Fiber-to-the-Curb scenarios are mainly influenced by fiber installation costs, the price reduction rate of photonic components may also affect the investment strategies. To speed up the cost reductions, telecom carriers should therefore invest on optical technology research and development. [ABSTRACT FROM AUTHOR]

The self-protecting multipath (SPM) is a simple and efficient end-to-end protection switching mechanism. It distributes traffic according to a path failure specific load balancing function over several disjoint paths and redistributes it if one of these paths fails. SPMs with optimal load balancing functions (oSPMs) are unnecessarily complex because traffic aggregates potentially need to be split which is an obstacle for the deployment of SPMs in practice. The contribution of this paper is the proposal of an integer SPM (iSPM), i.e., the load balancing functions take only 0/1 values and effectively become path selection functions. In addition, we propose a greedy heuristic to optimize the 0/1 distributions. Finally, we show that the iSPM is only little less efficient than the oSPM and that the computation time of the heuristic for the iSPM is clearly faster than the linear program solver for the oSPM such that the iSPM can be deployed in significantly larger networks. [ABSTRACT FROM AUTHOR]

We consider the problem of constructing a proxy-based overlay skeleton tree (POST) in the backbone service domain of a two-tier overlay multicast infrastructure. Spanning all multicast proxies deployed in the overlay backbone, POST acts as an efficient resource sharing platform for supporting large numbers of concurrent multicast sessions, without the need of tree computation for each individual session. The problem is concerned with deciding an appropriate deployment of multicast proxies in the overlay backbone, upon which we wish to find an optimal POST solution so that the maximum end-to-end latency is minimized subject to degree balancing constraints. This problem is shown to be NP-hard. We present a simple heuristic method for deploying multicast proxies, and devise a low complexity greedy algorithm for optimizing the end-to-end latency and degree distribution of POST. Simulation experiments confirm that our proposed approach yields good quality approximate solutions that are close to the optimum. [ABSTRACT FROM AUTHOR]

Parallel shared memory (PSM) routers represent an architectural approach for addressing the high memory bandwidth requirements dictated by output-queued switches. A fundamental related challenge pertains to the design of the high-speed memory management algorithm which is responsible for placing arriving packets into non-conflicting memories. In previous work, we have extended PSM results by introducing the concept of Fabric on a Chip (FoC). The latter advocates the consolidation of core packet switching functions on a single chip. This paper further develops the underlying technology for high-capacity FoC designs by incorporating a speedup factor coupled with a multiple packet placement process. This yields a substantial reduction in the overall memory requirements, paving the way for the implementation of large scale FoCs. We further provide analysis for establishing an upper bound on the sufficient number of memories along with a description of an 80Gbps switch implementation on an Altera Stratix II FPGA. [ABSTRACT FROM AUTHOR]