We investigated the packet delay and loss characteristics of the wide-area HIPPI-based CASA gigabit testbed. Blocking may occur because HIPPI is a connection-oriented point-to-point protocol that does not allow simultaneous connections. Blocking implies that data may not be delivered to a destination because of a competing (unrelated) concurrent connection. We show that HIPPI blocking can degrade performance by increasing delay and/or packet loss. In the CASA network under conditions of blocking, a tradeoff exists between packet loss and delay variance. The tradeoff point is determined by a combination of factors: source packet rate, mean blocking rate, and a configurable connection establishment timeout threshold.
We demonstrate that the delay/loss tradeoff manifests itself in TCP either by triggering the slow-start algorithm or inducing TCP to adjust retransmission timeout values due to increased delay variance. In the first case, where blocking triggers a slow-start, we have measured TCP throughput to drop by as much as 97% of its unblocked maximum. The latter case, retransmission/timeout adjustment, results in a 75% drop. The delay/loss tradeoff allows one to match network characteristics with application/transport requirements. With respect to TCP, high packet loss induces the slow-start policy even when the network may not be congested, leading to low throughput. In addition, the variance-sensitive TCP round trip time (RTT) estimator is sufficiently robust to avoid spurious retransmissions in HIPPI-based networks such as CASA.
Thus, the delay/loss tradeoff for TCP implies that avoiding packet loss is more important than minimizing delay. Other transport protocols or loss-tolerant applications may prefer a less reliable channel in exchange for bounded delay variance.
A paper detailing our observations and conclusions appeared in ``Usenix Symposium on High-Speed Networking, Aug 1-3, Oakland, CA. pp.45-59. [24]''