Trading Apps Optimized for High-Performance Hardware01.14.2013
Trading systems are being optimized to take advantage of advances in hardware, such as Intel-based and non-Intel-based multicore architectures.
“We are at the favorable intersection of two trends in low-latency trading,” said Mark Skalabrin, chief executive of Redline Trading Solutions, a provider of low-latency market data solutions. “Our customers—both buy side and sell side firms—are under relentless pressure to reduce their costs while at the same time bringing ultra low-latency performance to their trading applications.”
Redline’s third generation ticker plant, InRush 3, is optimized for Intel’s new multicore architecture. Customers can save on co-location costs with reduced server footprint and outsource exchange connectivity and infrastructure monitoring to Redline, Skalabrin said.
Current high-frequency trading platforms are typically implemented using software on computers with high-performance network adapters, such as field-programmable gate array (FPGA).
However, the downside of FPGA is that it is generally complicated and time consuming to set up, as well as to re-program, as the programmer has to translate an algorithm into the design of an electronic circuit and describe that design in specialized hardware description language.
“We have eliminated hardware acceleration, which had been in the product since 2008, and are relying on industry-standard Intel processors to do the job faster than FPGA,” said Skalabrin. “This has resulted in credible, reproducible and sustainable benchmarks, without the cost burdens and programming challenges associated with specialty hardware.”
The programming space on FPGA is also limited, so programs can’t be too big currently. Although, some tasks such as ‘circuit breakers’ are an ideal current use for FPGA technology.
A multicore processor is a single computing component with two or more independent actual central processing units (called ‘cores’). Processors are being designed with up to hundred cores.
Kx Systems, a provider of high-performance database and time series analysis, has optimized code in its kdb+ database to utilize processor-specific instructions available at run-time, including Intel’s Advanced Vector Extensions instructions, available on Intel’s latest generation of Sandy Bridge family of processors.
“It is vital to have software explicitly designed for multiple cores,” said Simon Garland, chief strategist at Kx Systems. “kdb+ is extraordinarily fast when benchmarking against other applications, as we are able to make full use of all available cores.”
Another type of processor, based on architecture developed by U.K. chip designer ARM Holdings, requires only 35,000 transistors, compared to the millions in many conventional processor chips, resulting in lower power usage.
“Because they’re smaller, you can have potentially thousands of ARM chips on a server,” said Garland. “Banks are interested in using ARM because the power demands are enormously lower than with conventional processors, which means they can expand their infrastructure with the same or fewer number of servers.”
Redline’s InRush 3 directly connects and manages multiple exchange feeds to high-performance client trading applications, and is optimized for platforms based on Intel’s Sandy Bridge processors in order to accelerate the tick-to-trade path.
In benchmark tests by research firm Securities Technology Analysis Center, InRush 3 produced tick-to-trade mean latency of 6.1 microseconds, maximum latency of 18.2 microseconds and standard deviation (jitter) of 1.4 microseconds.
Instinet Fox River Quant Solution will initially be rolled out in the Americas.
Instinet's Anushree Laturkar says investing in research and analysis is key to improving client outcomes.
The new platform launched via a partnership with BestEx Research.
It is the first liquidity-seeking algorithm on Atlas, the equities trading platform introduced in 2019.
Strategy Studio allows brokers to create their own execution platforms.