Research On The Fair Algorithms In Resilient Packet Ring Networks

In this paper, one new data transport technology named Resilient Packet Ring (RPR) in metropolitan area networks is discussed and anayzed in a great extend. Firstly, the background knowledge for RPR, the RPR protocol model and the RPR network/node architecture are fully introduced, then the operation mechanism of RPR and the key techniques in RPR networks are analyzed in detail. We do further and comprehensive research about the RPR fair algorithm in this paper. First we introduce several representative RPR fair algorithm which have been proposed before, the operation mechanism of each scheme is discussed and the performance comparion are made among them. After these we propose our three new RPR fair algorithms which can meet the key performance objectives of the RPR network, the simulation results also verify their satisfactory performance improvements.As a new network structure and data transport technology, the resilient packet ring (RPR), can overcome the drawbacks of the previous technology choices for MAN ring such as the SONET ring and the Gigabit Ethernet ring. RPR also adopts the dual-ring topology, spatial reuse, topology self-discovery, protect rotate and statistic multiplex etc, which enables it the ability to achieve the robust, high efficiency and economic simultaneously, and RPR has become the primary technology in metropolitan area networks. The IEEE 802.17 workgroup has complete the standard framework for RPR, however, many key technologies about RPR are in need of being further researched and discussed, one of which is the bandwidth fair allocation problem in RPR. How to allocate the ring bandwidth fairly as well as efficiently is an important issue in RPR networks and the RPR fair algorithms take the charge of such issue. Therefore the key technical challenge in RPR networks is the design of a RPR fair algorithm that can dynamically and simultaneously achieves the fairness, high utilization and spatial reuse. Before designing a RPR fair algorithm, one fairness principle, namely ring ingressaggregated with apatial aeuse (RIAS) should be considered first. The RIAS fairness has several key components, one of which is the level of the traffic granularity for fairness determination at a link is the so-called ingress-aggregated (IA) flow, which is the aggregate of all single flows originating form a given ingress node while endig at different destinations; the second component of RIAS is to ensure the maximal spatial reuse subject to the first constraint, that is, bandwidth on a link can be reclaimed by flows on it when it is unused either due to lack of demand or in cases of insufficient demand in which some flows crossing it are bottlenecked elsewhere; and the last component is the inner allocation of the same IA flow, that is, the bandwidth allocated to an IA flow should be allocated among its single flows according to some fairness principle.This paper discusses two kinds of bandwdith fair allocation modes in RPR networks, namely aggressive mode (AM) and conservation mode (CM) respectively, the common operation mechanism about the two modes is that a downstream congestion RPR node advertises a fair rate contained in the control message to the upstream nodes, which is to adjust the sending rates of the relevant nodes and deliver the congestion at the downstream. But both modes suffer from the serve oscillation, throughput degradation and too long delay etc, so that the performance requirements of RPR are not satisfied. Specifically, we also present three representative algorithms, namely Gandalf, Alladin and DVSR which have been proposed in RPR networks, the running mechanism of each scheme is analyzed respectively and the performance comparions are made among them.We propose in this paper three new RPR fair algorithms, namely ABA algorithm, CBFA algorithm and DBFA algorithm in detail, which are able to achieve the key performance objectives of RPR. Note these new algorithms do not the adopt the rate adjust mechanism used in AM or CM any more, on the contrary, we employ the classic control theory method...