A Throughput Study of RPR Networks Subject to Single Failures

Using typical operation scenarios, the throughput of IEEE 802.17 Resilient Packet Rings (RPR) operated in an SDH-like fashion is compared to the throughput of RPR with optimized bandwidth planning. Introduction IEEE 802.17 Resilient Packet Ring (RPR) has been introduced as a novel MAC layer protocol for ring based media [1]. RPR networks transport data (e.g., IP) packets over optical rings and are thus suited to interconnect routers at high bandwidth. An RPR network can support high and low priority client traffic and provide survivability upon failures. One motivation for RPR is bandwidth availability. Compared to protected SDH (or Sonet) rings, which use half of the provided bandwidth for working traffic and half of the bandwidth for protection traffic [2], RPR rings offer the full bandwidth capacity in failureless state. During a failure less bandwidth can be used in an RPR ring, however, there is no fixed bandwidth separation as in SDH. This paper addresses this issue using a case study for a typical operation of this type of ring. We compare the throughput of an RPR operated in an SDH-like fashion to an RPR with better managed bandwidth. Protection Mechanisms This section summarizes the protection methods from a network planning point of view. These aspects together with given scenarios are the basis for comparing different design alternatives in the next section. In the following we consider the general case of rings with N nodes. Links interconnecting the nodes of the ring are bidirectional (e.g., fiber-pairs). Each link has the same line bandwidth for both RPR and SDH protocols. Single link failures are assumed to be the most frequent failure to be considered for capacity planning. Capacity planning for single link failures includes protection capacity for single node failures, since the protection mechanisms for node failures are similar to link failures and since due to the failed node less traffic has to be recovered. In SDH, on each link half of the provided bandwidth is used for working traffic and half of the bandwidth is dedicated to protection traffic. In unidirectional SDH rings the sender copies the signal to both directions on a ring, such that in the failureless state the receiver can choose one of the incoming signals (Figure 1 a), for conciseness only one path is shown). Upon a failure on the chosen path the receiver can choose the backup path (Figure 1 c)). In bidirectional SDH rings there is during normal operation only one path between sender and receiver (Figure 1 a)). During a failure the failed segment is substituted by a protection path over the remaining ring (Figure 1 b)). It should be noted that the protection bandwidth of bidirectional rings, which is idle in the failureless state, can be used for pre-emptive traffic (in SDH terminology “low priority traffic”). As this involves additional administrative effort (configuration, traffic differentiation, bandwidth management, ...), we do not consider pre-emptive traffic in this paper. In RPR, on each operating link the entire provided bandwidth is usable. Two mechanisms have been proposed for protection switching: steering and wrapping. Using steering the sender chooses one of the two possible paths to reach the receiver (Figure 1 a)). Upon a failure on the chosen path the sender is notified about the failure and then switches to the other path (Figure 1 c)). Using wrapping, the behavior with respect to the protection paths is the same as in bidirectional SDH rings (Figure 1 a) and b)). Using a subsequent topology discovery and rerouting during the failure to avoid path loops (as between nodes C and D in Figure 1 b)) will result in a final routing as in the case of steering (Figure 1 c)) [3]. Figure 1: Protection path options b) and c) for a failure on link A-D of path a) Due to the fixed bandwidth allocation for working and protection, SDH rings offer the same bandwidth capacity in no and single failure cases. RPR rings offer twice the bandwidth of protected SDH rings in the failureless case. But as RPR rings reduce the bandwidth after protection switching, as a result appropriate automated bandwidth management and fairness algorithms can be activated [3]. However, during protection switching traffic loss should not become high, since in larger rings failure events become likely [4]. Therefore, failure events should be taken into account when planning the resources of an RPR ring. a) b) c) A