Dynamic resource sharing protection using label stacking and burst multiplexing in op
Active In SP
Joined: Jun 2010
12-10-2010, 12:18 PM
OBS.pdf (Size: 552.07 KB / Downloads: 50)
Dynamic resource sharing protection
using label stacking and burst multiplexing
in optical burst switched networks
Abstract: A protection and restoration mechanism is essential to provide reliable traffic transmission in networks, but it is not easy to apply existing protection mechanisms to optical burst-switching (OBS) networks, because of the unique properties of OBS. The authors introduce a 1:1 link-based protection that minimises burst losses by deflecting bursts until a source ingress router arbitrates a working path to a backup path when a link failure occurs. The authors also propose a genuine dynamic resource sharing (DRS) link protection in order to achieve high efficiency in network resource allocation and reliable protection simultaneously in OBS networks. DRS employs burst-multiplexing and label-stacking techniques in backup channel selection and always reserves an optimal number of backup channels according to the changes in traffic load in a working link and the quality of service (QoS) requirements of bursts. Simulation results show that the DRS can provide the same protection reliability with just half of the backup channels needed in a normal 1:1 protection in a light-load case, while guaranteeing the QoS requirements of class bursts. The authors show that this property also makes backup link sharing among a number of working links possible.
Optical burst switching (OBS) has increased the capability of cost-effective integrated services by statistical burst multiplexing, and rapid evolutions in optical devices and wavelength division multiplexing technology . However, much data loss can easily occur even because of a single link failure in the OBS network, as data bursts are transmitted by the one-way path reservation in an ingress edge router, and it is still difficult to deploy optical burst buffering in intermediate core routers. Until now, a few papers have dealt with this issue in terms of a protection and restoration in OBS networks . Oh et al. validated that their proposed recovery scheme has a high scalability to cope with a wavelength (or channel), link and node failures as well. However, a traditional circuitbased protection and restoration was applied to OBS networks without consideration about the unique property of the OBS, called tell-and-go. In the OBS, especially, a common just-enough-time (JET) scheme, the changes in a data path or offset time in a core router easily result in serious and contiguous burst losses. The OBS ingress node transmits a control packet in advance and reserves a burstswitching schedule based on the relative burst arrival time in every intermediate core router following the path. In the OBS ingress router, the offset time between the control packet and data burst is determined based on the end-toend path determined by a source routing and network load distribution, which includes a control packet processing time in each core router, the number of hops and quality of service (QoS) requirement levels. Thus, if unexpected changes occur in some parts of the path, bursts are discarded without exceptions until the ingress node handles that and controls burst transmission, even though bursts can be temporarily deflected in a local area. Thus, a single recovery easily results in contiguous burst losses in neighbouring links, because of the increase in the contentions of channel reservation. Bartos and Bhata proposed a revised protection signalling mechanism, but did not discuss how to adjust the source routing in the ingress edge node in correspondence to a local link protection and restoration. It is evident that the protection and restoration in the OBS network must be based on a link protection to minimise burst losses, because of the one-way tell-and-go transmission, and a path protection has to be applied in the ingress router, simultaneously. At the same time, in order to achieve high throughput, the statistical burst multiplexing property of the OBS has to be considered enough in resource reservation for backup links and paths. In this paper, we deal with the design issues of the link and path protection in the OBS networks, and propose a noble traffic load-based adaptive protection link reservation mechanism, called dynamic resource shared (DRS) protection. Based on monitored link status information, the DRS protection adjusts the number of protection channels according to the changes in the traffic load of the working link and QoS , and always reserves the optimal number of back channels by using burst multiplexing and label stacking. The DRS protection improves the performance in terms of the efficiency in resource allocation and channel utilisation, while guaranteeing the objective QoS level and protection reliability.