No abstract

Status of this Memo This document specifies an Internet standards track protocol for the Internet community, and requests discussion and suggestions for improvements. Please refer to the current edition of the "Internet Official Protocol Standards" (STD 1) for the standardization state and status of this protocol. Distribution of this memo is unlimited.

In this paper, we wish to derive analytic models that predict the performance of TCP flows between specified end-points using routinely observed network characteristics such as loss and delay. The ultimate goal of our approach is to convert network observables into representative user and application relevant performance metrics.The main contributions of this paper are in studying which network performance data sources are most reflective of session characteristics, and then in thoroughly investigating a new TCP model based on [1] that uses non-invasive network samples to predict the throughput of representative TCP flows between given end-points. This material is posted here with permission of the IEEE. Such permission of the IEEE does not in any way imply IEEE endorsement of any of the University of Pennsylvania's products or services. Internal or personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution must be obtained from the IEEE by writing to pubs-permissions@ieee.org. By choosing to view this document, you agree to all provisions of the copyright laws protecting it. Abstract-In this paper, we wish to derive analytic models that predict the performance of TCP flows between specified endpoints using routinely observed network characteristics such as loss and delay. The ultimate goal of our approach is to convert network observables into representative user and application relevant performance metrics. The main contributions of this paper are in studying which network performance data sources are most reflective of session characteristics, and then in thoroughly investigating a new TCP model based on [1] that uses non-invasive network samples to predict the throughput of representative TCP flows between given end-points.

One of the key issues in deployment of QoS is determining the set of applications or users, which are allowed to have a preferential access to network resources. The administrative criteria for regulating access to resources constitute the QoS policies. A policy could determine which of the reservation requests in the network be honored during the processing of a signaling protocol such as RSVP, or it could determine the class of applications or users which are to be placed in a specific DiffServ class of service. In this paper, we look at the issues that arise in the definition, deployment and management of policies related to QoS in an IP network. The paper provides an overview of the requirements for QoS policies, the alternate policy architectures that can be deployed in a network, the different protocols that can be used to exchange policy information, and the exchange of policy information among different administrative domains. We provide a coverage of the current issues being examined in IETF and other standard bodies, as well as issues explored in policy related research ongoing at different universities and research labs.

In this paper, we i n v estigate some issues related to the efcient provision of end-to-end delay guarantees in the context of the Guaranteed (G) Services framework [16]. First, we consider the impact of reshaping trac within the network on the end-to-end delay, the end-to-end jitter, as well as per-hop buer requirements. This leads us to examine a class of trac disciplines that use reshaping at each hop, namely rate-controlled disciplines. In this case, it is known that it is advantageous to use the Earliest Deadline First (EDF) scheduling policy at the link scheduler [8]. For this service discipline, we determine the appropriate values of the parameters that have to be exported, as specied in [16]. Subsequently, with the help of an example, we illustrate how the G service trac will typically underutilize the network, regardless of the scheduling policy used. We then dene a Guaranteed Rate (GR) service, that is synergetic with the G service framework and makes use of this unutilized bandwidth to provide rate guarantees to ows. We outline some of the details of the GR service and explain how it can be supported in conjunction with the G service in an ecient manner.

We introduce a queueing network model that allows us to capture the time-varying service delivered to a tra c stream due to the presence of random perturbations e.g. cross-tra c in a communication network. We rst present the model for a single queue and then describe how such queues may b e i n terconnected using the operations of fork or in-synchronization and join or out-synchronization. Such networks may be seen as a generalization of stochastic event graphs, and of the class of fork-join networks. The departure processes in such a network satisfy a system of equations along with the exogenous arrival processes and the service processes of the various queues. This system can be seen as a general time-varying linear system in the min; + semi-eld. We obtain an explicit representation of the departure process in terms of the exogenous arrival processes and the service processes. Su cient conditions are derived for this system to have a unique solution. We also study liveness and absence of explosion in this class of networks. Under appropriate stationarity and ergodicity assumptions, we establish stability theorems for such networks. To this end we rst obtain a rate result for the departure process in terms of the rates of the exogenous arrival process and the throughput of a saturated system. We then use this result to show that in a network with a single exogenous arrival, all queue lengths are nite with probability one if the arrival rate is less than the throughput of the saturated system, and to give a representation of the queue length process. This class of networks allows for a detailed description of controlled sessions in integrated service networks. We also show that it contains several earlier discrete event models of the literature pertaining to stochastic Petri nets, service curves and fork-join networks, and show h o w the present model uni es them in a single algebraic structure. This structure is that of a semiring of functions of two real variables, where the addition is the pointwise minimum and the multiplication a generalization of infconvolution.

Status of this MemoThis document specifies an Internet standards track protocol for the Internet community, and requests discussion and suggestions for improvements. Please refer to the current edition of the "Internet Official Protocol Standards" (STD 1) for the standardization state and status of this protocol. Distribution of this memo is unlimited.

In recent years, a number of link scheduling algorithms have been proposed that greatly improve upon traditional FIFO scheduling in being able to assure rate and delay bounds for individual sessions. However, they cannot be easily deployed in a backbone environment with thousands of sessions, as their complexity increases with the number of sessions. In this paper, we propose and analyze an approach that uses a simple buffer management scheme to provide rate guarantees to individual flows (or to a set of flows) multiplexed into a common FIFO queue. We establish the buffer allocation requirements to achieve these rate guarantees and study the trade-off between the achievable link utilization and the buffer size required with the proposed scheme. The aspect of fair access to excess bandwidth is also addressed, and its mapping onto a buffer allocation rule is investigated. Numerical examples are provided that illustrate the performance of the proposed schemes. Finally, a scalable architecture for QoS provisioning is presented that integrates the proposed buffer management scheme with WFQ scheduling that uses a small number of queues.