Affordable Access

A Quantitative Evaluation of Reordering in OBS Networks and its Impact on TCP Performance

Authors
Publisher
Other IFIP Publications
Publication Date

Abstract

A Quantitative Evaluation of Reordering in OBS Networks and its Impact on TCP Performance Jordi Perelló†, Sebastian Gunreben§ and Salvatore Spadaro† †Optical Communications Group (GCO), Signal Theory and Communications Department (TSC) Universitat Politècnica de Catalunya (UPC), Jordi Girona 1-3, 08034 Barcelona, Spain Tel: (+34) 93 401 7179, Fax: (+34) 93 401 7200 Email: [email protected], [email protected] §Institute of Communications Networks and Computer Engineering University of Stuttgart, Pfaffenwaldring 47, 70569 Stuttgart, Germany Tel: (+49) 711 685 67968, Fax: (+49) 711 685 57968 Email: [email protected] Abstract—This paper presents a layered framework to measure the reordering introduced by contention resolution strategies in OBS networks. In particular, characterization is based on the reordering metrics proposed by the IETF IPPM WG. The obtained results are twofold. First, they quantify the impact of burst reordering on TCP throughput performance. Second, they give insight into solving burst reordering by well dimensioned buffers. I. INTRODUCTION Optical Burst Switching (OBS, [1]) has emerged as a promising transport technology for next-generation Internet. As a matter of fact, OBS networks become a combination of packet and circuit switched networks, where packets are firstly aggregated in edge routers and, then, are sent as bursts along bufferless optical networks. This provides the benefit from statistical multiplexing in the optical domain, which allows better adaptation than circuit-switched networks (OCS) to higher layer dynamics. Besides, it lessens technology requirements compared to all-optical packet switching (OPS) networks. OBS relies either on one-way or two-way resource reser- vation. However, one-way resource reservation is usually im- plemented in wide area networks, where propagation delays may become much longer than burst transmission times. This means, that unlike in OCS networks, data transmission is not delayed until the recep

There are no comments yet on this publication. Be the first to share your thoughts.