The ability to quickly analyze timeseries of price and trade data is crucial for firms which trade financial instruments such as stocks, currencies, options and other derivatives. The growth of high-frequency trading makes it important for firms to rapidly store, retrieve and analyze massive quantities of data in order to back-test new trading algorithms, monitor their risk, and satisfy regulatory demands. At the same time, surging data volumes brought about by computerized trading make these tasks increasingly challenging in markets which can 'tick' (publish changes to prices) millions of times per second.
Peter Lankford, founder & director of STAC, said: "We're pleased that Kx and Oracle have taken the lead in publishing STAC-M3 Benchmarks. This testing responds to a direct need for performance data expressed by some of the world's leading financial institutions. We look forward to further benchmark results from Kx and Oracle, as well as other vendors of software and hardware that can be used for managing large timeseries of data."
Simon Garland, Kx chief strategist, said: âWe have worked closely with STAC and Oracle to create benchmarks. Going forward, the benchmarks will give financial institutions clearly defined criteria for measuring the performance of their systems. When setting up large, critical databases people need to be sure that the configuration they use is delivering the best possible performance within their hardware and software constraints. The problem is nobody has the time and resources to do an in-depth evaluation of all the alternatives and ideally to regularly revisit the chosen solution. These new benchmarks from STAC are valuable components of these decisions."
This series of tests is designed to set a standard for future benchmarks to be carried out by STAC in conjunction with various vendors. The results are vital for financial institutions needing to address their CEP requirements, facing the pressures of constantly increasing data volumes and regulatory demands. The STAC-M3 Benchmark suite is the first benchmark standard for market data timeseries (tick) databases, allowing vendors to measure their performance and financial institutions to test their own systems.
The STAC-M3 Benchmark standards will accelerate market selection of new technologies by enabling user firms and vendors to measure the performance of their database solutions in a common way. Trading firms will be able to quickly assess the benefit of innovations at all layers of the tick-database âstackâ including application software, operating systems, processors, servers, and storage architectures.
Amir Halfon, Oracleâs senior director of financial services technology, added: âOracleâs unique storage server, the Sun Fire X4540, allowed us to achieve extreme performance with Kx by placing the data as close to the CPUs as possible. The Sun Fire X4540 provides fast, high throughput access between the threads of execution and the storage subsystem, offering up to 96TB of storage in a compact, cost-effective package. The tight integration between compute and storage within the Sun Fire X4540 eliminates the need for an external SAN and provides an ideal platform for high-performance analytics solutions such as Kx kdb+.â
The STAC-verified benchmarks were carried out using Kx kdb+ v2.6 on the Oracle Sun Fire X4540 storage server using version 0.92 of the STAC-M3 Benchmark specifications. STAC-M3 v0.92 has been submitted to the full Council for ratification, and a vote is pending.