Logo image
一個在差別式服務環境中提供嚴格服務品質保證之頻寬預留協定
Thesis

一個在差別式服務環境中提供嚴格服務品質保證之頻寬預留協定

林俊甫
Masters, National Tsing Hua University
1999

Abstract

差別式服務整合式服務頻寬預留協定 Internet QoSDiff-ServInt-ServBRP
In traditional IP-based network, all the Internet traffic is delivered in best-effort manner without any QoS guarantee. As Internet multimedia applications become popular, current network framework cannot provide enough quality of service to meet the demand for a variety of new applications. In order to solve the problem, Internet Engineering Task Force (IETF) has proposed two network frameworks successively: Integrated Services (Int-Serv) and Differentiated Services (Diff-Serv). Resource Reservation Protocol (RSVP) is employed in Int-Serv to make the reservation along routing path before traffic is delivered. Diff-Serv specifies different forwarding treatments to provide a number of services corresponding to different QoS provisions. The Expedited Forwarding per-hop behavior (EF PHB) is specified to provide a low loss, low latency, assured bandwidth and end-to-end service through Diff-Serv domains. The EF PHB is implemented by configuring nodes such that the EF aggregate has a minimum departure rate independent of other traffic at the node. However, only the rate configuration of EF PHB is not able to be a hard QoS guaranteed service which must experience the strict constraint of QoS specification. In this thesis, we aim to provide a hard QoS guaranteed service with end-to-end behavior across domains in Diff-Serv. We add bandwidth reservation process for each EF session in Diff-Serv with strict QoS provision. For this purpose, we utilize three mechanisms for the new framework: (1) local centralized resource management, (2) centralized route discovery, (3) a sender-initialized signaling protocol. The signaling protocol we proposed is named Bandwidth Reservation Protocol (BRP) and it is responsible for exchanging messages between bandwidth broker and the edge router. By simulation, we evaluate the performance of our proposal against current Diff-Serv framework. Our approach has excellent performance on congestion probability and average delay for both short-term and long-term EF traffic.

Metrics

1 Record Views

Details

Logo image