End-to-end failover thresholds for transport layer multihoming
Armando Caro, Paul D. Amer, Randall Stewart · 2005
SCTP's multihoming failure detection time depends on three tunable parameters: RTO.min (minimum retransmission timeout), RTO.max (maximum retransmission timeout), and Path.Max.Retrans (threshold number of consecutive timeouts that must be exceeded to detect failure). RFC2960 recommends Path.Max.Retrans = 5, which translates to a failure detection time of at least 63 seconds - unacceptable to many applications. This research investigates the tradeoff between a more aggressive (i.e., lower) threshold, and spurious failovers for the application of bulk file transfer. We surprisingly find that spurious failovers do not degrade overall performance, and sometimes actually improve goodput performance.