The Securities Technology Analysis Center (STAC), a developer of standard benchmarks for trading technology, today cited IBM as the first vendor to complete testing against their STAC-M2 Benchmark specification, which evaluates the ability of messaging middleware to handle real-time messaging in a variety of customer defined test conditions.
The STAC-M2 tests measure key metrics such as latency, throughput, power efficiency, and CPU/memory consumption in multiple scenarios. The messaging stack tested consisted of IBM WebSphere MQ Low Latency Messaging, Red Hat Enterprise Linux, quad-data rate InfiniBand from Voltaire and Mellanox, and IBM blade servers using Intel X5570 "Nehalem" processors.
According to the STAC-M2 Report, in use cases where latency, or speed, is a key differentiator, IBM's MQ Low Latency Messaging software achieved average latency as low as eight microseconds at 50 thousand messages per second, and 12 microseconds at one million messages per second. The report further states that the 99th percentile latency was at most 18 microseconds in these same tests, and standard deviation, a measure of latency jitter, did not exceed two microseconds. All tests were executed with message sizes representing customer defined payloads of approximately 270 bytes, including STAC imposed message headers. While these tests stress certain deployment scenarios, there are others where IBM has measured considerably higher throughput and lower latency, according to an IBM press release.
The emphasis on speed, capacity and efficiency is causing dramatic changes within the financial markets. Electronic trading volumes are growing dramatically, placing intense pressure on the existing IT infrastructure to deliver the speed and capacity needed to keep up with an unprecedented flow of data. For example, automated trading by high-frequency trading firms now makes up 73 percent of all US equity trading volume. WebSphere MQ Low Latency Messaging provides a messaging transport that is optimized for the very high-volume, low-latency requirements of financial markets firms.
"The STAC test results reinforce the consistent high performance characteristics necessary to serve as an underlying transport in varying use-cases, including direct exchange connectivity in products like WebSphere Front Office for Financial Markets," said Guy Tagliavia of IBM, in a statement. "These products together form a powerful platform for market data integration, complex event analysis, transaction processing, surveillance, risk management, and customer integration for the evolving demands of the securities trading industry." The latest release of IBM MQ Low Latency Messaging offers new features such as assured message delivery, high-reliability, self-management, and a technology preview of patented message-based clock synchronization. These features assure that messages are made persistent before delivery, systems can be deployed in multi-tiered environments, applications can be automatically configured, and end-to-end latency can be monitored, according to IBM.
WebSphere MQ Low Latency Messaging is part of the MQ family of connectivity middleware, offering a range of messaging capabilities. MQ is currently used by the top 20 banks in the world and more than half of the Global 500 banks worldwide, according to IBM.
Also announced this week, IBM WebSphere Front Office for Financial Markets v3.0.2 includes a 76 percent decrease in latency and a 40 percent increase in throughput, new support for additional direct exchange feeds covering Japan, Brazil, and Mexico, and a new feed handler toolkit that will allow clients to build their own custom feed handlers, according to IBM.
In addition to support for IBM's Nehalem-based System x, both of these new WebSphere product releases also add support for IBM's Power-based System p running Linux, and allow clients to gain competitive advantage and optimize their IT investments. By running their applications and middleware on IBM's Power Systems, clients can achieve cost savings through server consolidation for reduced data center space and power consumption, increased availability and improved service through virtualization capabilities, claims IBM.