SpryWare, a premier provider of Ultra Low Latency feed handlers and direct market data technology, today announced the release of their third generation ticker plant software taking flexibility, throughput, and low latency to new levels.
The cutting edge ticker plant technology is designed with an impressive list of enhancements including the option to install the software as an Embedded MIS (Embedded Market Information Server), deploying it as a full In-Process solution, bringing the feed handlers directly into the client trading application on either a Linux or Microsoft Windows server environment. With the In-Process configuration, latency is brought down to single digit microsecond level on key market data feeds.
"Our software based solution allows us to fine tune The Market Information Server (MISTM) with any of the latest product releases from companies like Intel and SolarFlare" says Claus Jensen, Director of Marketing at SpryWare. "As an example The MIS has been optimized to take full advantage of the complete 64 bit instruction set and all the features such as Simultaneous Multi-Threading (SMT) offered by the latest Intel Xeon processor 5600 series. With that we have been able to see a throughput increase of up to 61% compared to the previous processor generation" continues Mr. Jensen.
The parallel processing power of the MIS has been further optimized and is now capable of processing rates of up to 750,000 messages per second per core or 375,000 messages per second per processor thread.
With SolarFlare, SpryWare has tested and certified the MIS using OpenOnload with the Solarstorm network adapter. OpenOnload is a high performance network stack for Linux. It performs network processing at the user-level, bypassing the OS kernel thus reducing networking overhead and latency.
"The second generation of the SpryWare MIS was all about moving from milliseconds to microseconds, and that was all about removing resource contention and streamlining parallel processing across multiple cores. Now it is all about moving from microseconds to nanoseconds, which is more about processor cache line management and exploiting specific CPU instruction sets. We really are countre counting every CPU cycle. This gives us not only a reduction in latency by a full order of magnitude, but equally increases message processing capacity" says Daniel May, Director at SpryWare. "We delivered generation three with little or no changes to the existing SpryWare API interface, so a trading application can move from a remote Internet desktop, to the cloud, to the data center, all without a recompile. It just works. This gives a tremendous number of deployment options for a customer."