Performance

The Wombat feed handlers were originally designed to support options market making systems utilizing the high volume OPRA feeds. The resulting architecture is now an industry leader in three key competitive areas in the emerging market:

• Ultra-low latency
• Scalability and throughput
• CPU efficiency

First and foremost the Wombat feed handler architecture is designed to provide a blisteringly fast ticker-plant. The technology makes use of "vertical" threads. In layman's terms this means that a particular incoming packet is processed on a single thread of execution. The net result is that processing is not interrupted from the time a message arrives on the incoming feeds to the time its derivatives are broadcast onto the client network. This gives an end to end ticker plant latency of between 0.2 and 0.4 ms, sometimes lower.

Scalability is the second key requirement underpinning the Wombat feed handler architecture. This is achieved in a fully distributed architecture.

Once again the distributed architecture stemmed from working with the OPRA feed. Even in the late 1990s it was clear that OPRA market data volumes were going to increase dramatically. The feed itself supported scalability and increasing volumes with multiple sub feeds. The original Wombat architecture allowed the ticker plant to scale sideways by subdividing the data across multiple instances of the feed handler.

This distributed architecture is discussed in more detail in the "Wombat Multicast Feed Handlers" white paper downloadable from the Technical / Whitepapers section of the web site.