Solarflare, Arista Complete Multi-Vendor Benchmarking Test

Arista Networks helped Solarflare to successfully complete a multi-vendor test using the STAC-M2 Benchmark to measure the messaging performance of the solution stack.

STAC-M2 Benchmark was established to measure the ability of a solution to handle real-time market data in a variety of configurations.

The companies note that the multi-vendor testing achieved very low mean latencies with very small standard deviations of the latency, also known as jitter.

The specific results may not be disclosed publicly as per the policies of STAC. The full report is offered to banks, trading firms and other end-user members of the STAC Benchmark Council with access to the STAC Vault.

According to a release, the solution stack for this report used products available for shipping to financial firms today and consisted of the following components:

IBM (News - Alert) WebSphere MQ Low Latency Messaging v2.4.0.2

Red Hat Enterprise Linux Server AS 5.5

Intel Xeon CPU E5530 @ 2.40GHz

IBM x3550 M2 Model 7946-E2U

Solarflare (News - Alert) SFN5122F 10GbE adapter with OpenOnload

Arista Networks 7124S 10GbE switch

Readers of the report will know the very low standard deviation of latency (jitter) that was observed during testing.

Commenting on the test Solarflare says that for the purpose, Solarflare's OpenOnload application acceleration middleware was used in combination with the SFN5122F to enable full operating system bypass.

This significantly reduces host processing overheads, enables high transaction rates and significantly reduces application latency with very low jitter.

Configuration of Solarflare's OpenOnload library enabled the application threads to run without contention with each other, and without requiring any interrupt processing or other interaction with the kernel active on the dedicated core.

The Arista 7124S switch features ultra-low latency and cut-through switching functionality. The company claims that these features and the switches’ dynamic allocation of packet memory are major contributors to the benchmark results.

Read a related article at TMCnet: “As FASER Energy Benchmarking Software Phases Out, EnerCop from Kilojolts Fills the Void.”