19. Latency Measurement Overview

In our work at 29West, we wanted to measure the latency involved in delivering messages through our LBM product. Latency in messaging systems with LBM must often be measured in microseconds because sub-millisecond results are common. The extremely low latency of LBM raises measurement issues not often found with leading messaging systems. Following sections discuss some of these issues and provide results from some of our latency measurements. See Section 17 for background on the latency sources we've noted during testing in our Latency Busters® Lab.

One of the largest sources of latency we saw in our tests was network loss repair. Time is required for the transport layer to first detect the loss, request retransmission, and finally for the repair to arrive. See Section 17.3 for a general discussion of retransmissions and latency. See Section 2 for a discussion of such latency in TCP. For reliable multicast transport protocols such as 29West's LBT-RM, minimum loss repair latency comes at a cost of efficiency. Systems using reliable multicast protocols for their scalability often have to trade off efficiency to get the lowest possible loss repair latency.

No matter what transport protocol is in use, it's important to check for loss at the first hint of latency. See the end of Section 2.2 for some advice on diagnosing TCP loss. See Section 15.3 for advice on how to avoid loss when using multicast.