Specifications

18
High Performance Trading/Algo Speed with Wombat Design and Implementation Guide
OL-15617-01
Testing
Figure 9 shows a histogram of the mean latency observations for both Cisco IB and Ethernet.
Figure 9 Transport Mean Latency at 1x Recorded Rate
Because the update rate varied over time, it is important to check whether there is a correlation between
update rate and latency. Live market data does not flow at a steady state. It has peaks and troughs and
bursts of updates over small and large intervals. Yet many latency-sensitive applications value
predictability of latency. The less variable a system’s latency is as traffic ebbs and flows, the more
beneficial it is.
Figure 10 plots updates per second (per line/client) against mean latency. Ethernet exhibited decreasing
mean latency with increasing update rate. Note that this is not the ingress update rate (i.e., the rate at
which arrived at the OPRA FH machine and the Wombat tools do not measure). It is the rate at which
updates arrived at the mamaperf consumer. This makes it more difficult to draw conclusions about cause
and effect, but it is likely due to batching and flushing behavior that becomes more efficient with higher
update rates that InfiniBand does not exhibit the same behavior is consistent with this explanation, since
there is much less buffering.