Abstract

We consider the problem of allocating bandwidth to competing flows in an MPLS network, subject to constraints on fairness, efficiency, and administrative complexity. The aggregate traffic between a source and a destination, called a flow, is mapped to label switched paths (LSPs) across the network. Each flow is assigned a preferred ('primary') LSP, but traffic may be sent to other ('secondary') LSPs. Within this context, we define objectives for traffic engineering, such as fairness, efficiency, and preferred flow assignment to the primary LSP of a flow ('Primary Path First', PPF). We propose a distributed, feedback-based multipath routing algorithm that attempts to apply additive-increase and multiplicative-decrease (AIMD) to implement our traffic engineering objectives. The new algorithm is referred to as multipath-AIMD. We use ns-2 simulations to illustrate the fairness criteria and PPF property of our multipath-AIMD scheme in an MPLS network. Document type: Part of book or chapter of book

Abstract

In this paper, to improve the burst loss performance, we actively avoid contentions by proposing a novel congestion control scheme that operates based on the highest (called peak load) of the loads of all links over the path between each pair of ingress and egress nodes in an Optical Burst Switching (OBS) network. Document type: Part of book or chapter of book

Abstract

Part 1: Autonomic and Decentralized Management International audience Energy efficiency has become a major research topic in the Internet community as a result of unprecedented rise in the Information and Communication Technology (ICT) sector. One typical approach towards energy efficiency is to select a subset of IP routers or interfaces that will go to sleep mode during the off-peak period. However, on-the-fly network reconfiguration is generally deemed harmful especially to real time packets due to routing re-convergence. In this paper, we develop an efficient algorithm for achieving energy efficiency which is disruption free. The objective is to incrementally wake up sleeping links upon the detection of increased traffic demand. Unlike normal approaches of manipulating link weights or reverting to full topology in case of even minor network congestion and thereby sacrificing energy savings, our algorithm wakes up the minimum number of sleeping links to the network in order to handle this dynamicity. The performance of our algorithm was evaluated using the GEANT network topology and its traffic traces over a period of one week. According to our simulation results, up to almost 47% energy gains can be achieved without any obstruction to the network performance. Secondly, we show that the activation of a small number of sleeping links is still sufficient to cope with the observed traffic surge. Document type: Part of book or chapter of book

Abstract

This paper presents a prototype for single-rate reliable multicast congestion control, which has been built into an existing commercial whiteboard. The prototype was developed using a novel scheme that was engineered around conflicting industry provided requirements for collaborative workspaces. This required the scheme to be both low-weight when used with many senders and compatible with NAT, firewalls and reflectors. The key to overcome this conflict was to combine congestion control and recovery feedback. This differs from many current solutions in that they are often designed for use with a wide variety of protocols and thus operate independent of the recovery mechanism. This paper does not go into the detail required to specify a protocol but instead discusses a few important design requirements for multi-sender applications, which are generally not considered by current research, and describes an approach towards meeting these requirements. Document type: Part of book or chapter of book

Abstract

Most congestion control algorithms try to emulate processor sharing (PS) by giving each competing flow an equal share of a bottleneck link. This approach leads to fairness, and prevents long flows from hogging resources. For example, if a set of flows with the same round trip time share a bottleneck link, TCP’s congestion control mechanism tries to achieve PS; so do most of the proposed alternatives, such as eXplicit Control Protocol (XCP). But although they emulate PS well in a static scenario when all flows are long-lived, they do not come close to PS when new flows arrive randomly and have a finite amount of data to send, as is the case in today’s Internet. Typically, flows take an order of magnitude longer to complete with TCP or XCP than with PS, suggesting large room for improvement. And so in this paper, we explore how a new congestion control algorithm — Rate Control Protocol (RCP) — comes much closer to emulating PS over a broad range of operating conditions. In RCP, a router assigns a single rate to all flows that pass through it. The router does not keep flow-state, and does no per-packet calculations. Yet we are able to show that under a wide range of traffic characteristics and network conditions, RCP’s performance is very close to ideal processor sharing. Document type: Part of book or chapter of book

Abstract

We address the logical topology design problem (LTD) in WDM transport networks under static traffic assumptions. We start with one of the standard MILP formulations of the LTD problem that aims at optimizing the network congestion. We propose an improvement to this model that additionally optimizes the average hop count. We then derive a new MILP model that compels the traffic to be atomically routed. This last model enables fair comparisons of solutions obtained with MILP formulations and with metaheuristic algorithms. The latter allow us to deal with large size networks whereas the former are limited by their computational complexity. In this paper Tabu Search is used to tackle the LTD problem. We compare and discuss the logical topologies computed by the various methods described in the paper. Document type: Part of book or chapter of book

Abstract

Widespread deployment of multicast depends on the existence of congestion control protocols that are provably fair to unicast traffic. In this work, we present an optimization-based congestion control mechanism for single-rate multicast communication with provable fairness properties. The optimization-based approach attempts to find an allocation of rates that maximizes the aggregate utility of the network. We show that the utility of multicast sessions must be carefully defined if a widely accepted property of aggregate utility is to hold. Our definition of session utility amounts to maximizing a weighted sum of simple utility functions, with weights determined by the number of receivers. The fairness properties of the optimal rate allocation depend both on the weights and form of utility function used. We present analysis for idealized topologies showing that while our mechanism is not strictly fair to unicast, its unfairness can be controlled by appropriate choices of parameters. Document type: Part of book or chapter of book

Abstract

The purpose of this tutorial presentation is to introduce G-Networks, or Gelenbe Networks, which are product form queueing networks which include normal or positive customers, as well as negative customers which destroy other customers, and triggers which displace other customers from one queue to another. We derive the balance equations for these models in the context of multiple customer classes, show the product form results, and exhibit the traffic equations which - in this case, contrary to BCMP and Jackson networks - are non-linear. This leads to interesting issues of existence and uniqueness of the steady-state solution. Gelenbe Network can be used to model large scale computer systems and networks in which signaling functions represented by negative customers and triggers are used to achieve flow and congestion control. Document type: Part of book or chapter of book

Abstract

MaxNet is a distributed congestion control architecture in which only the most severely bottlenecked link on the end-to-end path controls the source rate. This paper shows that the small-signal convergence speed of MaxNet is higher than that of conventional architectures, such as the current Internet or REM. It also shows that MaxNet decouples the control, so that each pole position depends only on parameters of one bottleneck link and of the sources controlled by that bottleneck, enabling optimal pole placement. Document type: Part of book or chapter of book

Abstract

A TCP flow is congestion responsive because it reduces its send window upon the appearance of congestion. An aggregate of nonpersistent TCP flows, however, may not be congestion responsive, depending on whether the flow (or session) arrival process reacts to congestion or not. In this paper, we describe a methodology for the passive estimation of traffic congestion responsiveness. The methodology aims to classify every TCP session as either "open-loop" or "closed-loop". In the closed-loop model, the arrival of a session depends on the completion of the previous session from the same user. When the network is congested, the arrival of a new session from that user is delayed. On the other hand, in the openloop model, TCP sessions arrive independently of previous sessions from the same user. The aggregate traffic that the open-loop model generates is not congestion responsive, despite the fact that each individual flow in the aggregate is congestion responsive. Our measurements at a dozen of access and core links show that more than 60-80% of the traffic that we could analyze (mostly HTTP traffic) follows the closed-loop model. Document type: Part of book or chapter of book