Supporting synchronous and asynchronous traffic in optical-burst-switched metro networks
Sheeshia, Sami F
MetadataShow full item record
We investigate the role that Optical Burst Switching (OBS) can play in supporting asynchronous (data) traffic and synchronous (voice) traffic over DWDM. Specifically, we focus our study on Metropolitan Area Networks (MANs), which have traditionally been constructed using the Synchronous Optical NETwork (SONET) transport. This research presents new concepts and modeling methods that can be used to enhance and transform OBS into an efficient transport for MANs. We develop analytical and simulation models to evaluate OBS in supporting asynchronous traffic e.g. Ethernet traffic, and compare its effectiveness with those of SONET and Next-Generation SONET (NG-SONET) in MANs. Results show that OBS is more efficient and data-friendly than SONET/NG-SONET when supporting bursty Ethernet traffic. We then develop the analytical and simulation models for the delay and loss rate. Our results show that OBS yields semi-constant metro end-to-end delay and delay variance that are lower than those of SONET/NG-SONET. On the other hand, SONET/NG-SONET achieves lower loss rate than OBS with the distributed processing scheme (OBS-JET) and higher loss rate than OBS with the centralized processing scheme (WR-OBS). For OBS to be viable as a metro transport, it must support synchronous traffic such as voice services. We introduce the Synchronous Optical burst Switching (SOBS) framework and describe the challenges of supporting the synchronous and asynchronous traffic in the same OBS network. SOBS extends OBS by using a two-way path setup signaling scheme for guaranteed delivery, and a pro-active periodic reservation technique that does not require time synchronization or a significant amount of signaling. We then introduce the concept of burst grooming as a means of switching small synchronous and asynchronous bursts without adversely affecting the node efficiency. We differentiate between different implementations and develop related module architectures and equivalent network models. An analytical model for the throughput is then derived and the grooming problem is formulated as a set of optimization routines whose objective is to maximize the traffic throughput when the traffic delay is bounded or requires minimization. Results demonstrate that burst grooming yields throughput gains for both synchronous and asynchronous traffic at the expense of small delays.