Scalable Multi-Failure Fast Failover via Forwarding Table Compression

In datacenter networks, link and switch failures are a common occurrence. Although most of these failures do not disconnect the underlying topology, they do cause routing failures, disrupting communications between some hosts. Unfortunately, current 1 : 1 redundancy groups are only partly effective at reducing the impact of these routing failures. In principle, local fast failover schemes, such as OpenFlow fast failover groups, could reduce the impact by preinstalling backup routes that protect against multiple simultaneous failures. However, providing a sufficient number of backup routes within the available space provided by the forwarding tables of datacenter switches is challenging. To solve this problem, we contribute a new forwarding table compression algorithm. Further, we introduce the concept of compression-aware routing to improve the achieved compression ratio. Lastly, we have created Plinko, a new forwarding model that is designed to have more easily compressible forwarding tables. All optimizations combined, we often saw compression ratios ranging from 2.10x to 19.29x.