Specifications

22
High Performance Trading/Algo Speed with Wombat Design and Implementation Guide
OL-15617-01
Testing
Figure 12 plots the updates per second (per OPRA line/client) against mean latency. Unlike at 1x
recorded rate, the Ethernet latencies do not fall appreciably with update rate. This is consistent with the
previous hypothesis (that the inverse relationship was due to batching and flushing). At the rates
experienced in the 4x scenario, the buffering and flushing may have been maxed out.
Figure 12 Latency vs Update Rate at 4x Recorded Rate
Latency Dispersion
Latency dispersion—or roughly, how spread out the latency values are—is just as important as the
average latency. Like the man who drowned in a river that was six inches deep on average, trader’s do
not care a lot about mean latency if the quote or order they cared the most about was delayed far beyond
the mean. Moreover, if mean latency is low enough, reducing dispersion can actually be more important
than reducing mean latency, since dispersion befuddles trading algorithms with unpredictability.
Dispersion is an abstract concept that is captured in a variety of statistics. Below, we look at standard
deviation and maximum latencies.