文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > BPR + A PACKET SCHEDULING ALGORITHM FOR PROPORTIONAL DELAY DIFFERENTIATION
侵权投诉

BPR + A PACKET SCHEDULING ALGORITHM FOR PROPORTIONAL DELAY DIFFERENTIATION

Abstract- Proportional service differentiation models are simple and effective approaches to service differentiation in Internet. Schedulers for delay differentiation are important elements of these models. We describe a new scheduler for proportional dela

BPR+: A PACKET SCHEDULING ALGORITHM FOR PROPORTIONAL

DELAY DIFFERENTIATION

Vladimir Vukadinovi 1, Grozdan Petrovi 2, Ljiljana Trajkovi 1

1

2

School of Engineering Science, Simon Fraser University, Vancouver, Canada

Faculty of Electrical Engineering, University of Belgrade, Belgrade, Serbia and Montenegro

Abstract - Proportional service differentiation models spacing between traffic classes, such as delay and loss are simple and effective approaches to service rate ratios. differentiation in Internet. Schedulers for delay

differentiation are important elements of these models.

We describe a new scheduler for proportional delay

differentiation called BPR+: Optimized Backlog-Proportional Rate scheduler. BPR+ considers waiting

times of packets that are backlogged in queues in order to make precise scheduling decision. In this paper, we evaluate the performance of BPR+, which is achieved by taking into account complete information regarding backlogged packets. 1. INTRODUCTION

Internet applications have diverse service requirements. Real time applications, such as IP telephony and video on demand, require low delay and delay jitter while being relatively insensitive to packet losses. Applications dealing with bulk data transfer, such as File Transfer Protocol (FTP) and Simple Mail Transfer Protocol (SMTP), require sufficient throughput and reliability while being tolerant to network delay. Today’s best-effort Internet service is not adequate for many applications and

users. Considerable Internet research has been devoted

recently to providing different levels of service to

applications.

Initial proposals for service differentiation, such as Integrated Services (IntServ) model [1], provided strong service guarantees, with absolute per-flow bounds on delays, loss rates, and throughput. They required substantial modification to current Internet architecture. Furthermore, IntServ model raised scalability concerns because maintaining per-flow state imposes large

computational and memory requirements on core routers

in Internet. The interest has shifted recently to class-based

service architectures that provide relative service

guarantees to flow aggregates called classes: flows

belonging to a higher priority class receive better

performance compared to those of a lower priority class.

A service model known as Differentiated Services (DiffServ) [2] has also received considerable attention.

Unlike in IntServ model, DiffServ routers employ per-class buffer management and packet scheduling, where

the number of classes is significantly lower than the number of flows. There are two approaches for delivering QoS guarantees in DiffServ: absolute and relative service differentiation. Relative service differentiation is simpler and easier for deployment compared to absolute

differentiation. Network administrator may vary quality

Recently proposed Proportional Differentiated Services

(PDS) model [3], [4] defines a service model where the

ratios of loss rates and packet delays between successive

priority classes remain approximately constant.

Scheduling and buffer management algorithms have to be designed in order to support relative service differentiation. Packet scheduling specifies the packet serving discipline in a network node. Incoming packets are organized by the buffer management algorithm into

different logical queues. After a packet has been transmitted from the buffer, the packet scheduling algorithm decides which queue should be served next.

Since QoS parameters of a traffic flow are significantly affected by the choice of scheduling mechanism, considerable research efforts have been focused on designing various scheduling schemes for relative delay differentiation [3]-[10]. However, these schemes proved to be efficient only under certain conditions (heavy traffic load, sufficiently large buffers) and only on certain timescales (short or long). In this paper, we describe a new scheduling mechanism

that aims to provide more precise and consistent relative

delay differentiation between real-time and non-real-time

(elastic) traffic. The new mechanism called Optimized

Backlog-Proportional Rate scheduler (BPR+proportional delay differentiation. BPR+

) provides

considers waiting times of all packets that are backlogged in queues in order to make a scheduling decision. We searched for an “optimal” scheduler for proportional delay differentiation, even if its implementation may be to complex. The remainder of the paper is organized as follows: a

brief description of existing scheduling mechanisms for

proportional delay differentiation is given in Section 2.

The Optimized Backlog-Proportional Rate scheduler

(BPR+) is described in Section 3. In Section 4, we

evaluate the performance of the proposed scheme using

various simulation scenarios. Conclusions are presented

in Section 5. 2. RELATED WORK Several schedulers for relative delay differentiation have been recently proposed. We review here two schedulers: Waiting Time Priority (WTP) and Backlog-Proportional Rate (BPR) schedulers [3], [4]. WTP scheduler is based on Kleinrock’s Time-Dependent-Priorities (TDP) algorithm [11]. Priority of a packet from

第1页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email