The ‘Sustained FIX message rate - IP stats + wiretime latency’ benchmark measures VMX-Capture stack probe performance for FIX data. This benchmark is measured in messages per second.

The FIX data that is used for the test are streaming executable prices for foreign exchange, but it is a generic decoder that is used for all TCP-based FIX messaging (i.e. the same decoder is used for FIX order processing).

The predominant message type in the capture is the Market Data Snapshot Full Refresh (35=W), with 6 Market Data Entries (268=6). The average FIX message length is 352 bytes.

This benchmark is for the capability to decode and produce aggregated statistics.

It does not measure the ability of the system to correlate messages with each other (for example, to produce a request/response measurement like an order-to-ack acknowledgment measurement).

It does not measure the ability to persist the individual decoded order messages to a structured database (for example, to perform an order search), however it does include the capability to write all of these messages as packet captures, and the Beeks Analytics system provides the capability to retrospectively interrogate these packet captures. See the Prism section of the Analytics Concepts Guide for more information.

What statistics are available to me with an ‘IP stats + wiretime latency’ configuration for FIX messaging?

As per the ‘IP stats only’ for market data (see Stack Probe performance for Market Data ) but also including:

  • Messages per second

  • App-to-wire latency (also known as Wire time)

App to wire latency is the time that Beeks Analytics calculates between the SendingTime timestamp written within the message and the time it is observed at the Visibility Point (also known as the Wire time). 

The figure can only be as accurate as the time resolution for the SendingTime field within the message (FIX tag 52 or equivalent field in other trading protocols). Often this is recorded to the nearest millisecond, whereas the wire time is recorded to the nearest microsecond. As a result, figures of hundreds of microsecond can be considered normal.

The app-to-wire latency can be reported at different user-definable percentiles (e.g. 95th percentile, 99th percentile, as well as minimum, mean and maximum).