Sifma Technology Management Conference
Into the Sifma Era | AML Software Landscape Consolidates and Expands | People Are Scarce Again | ATSs in Europe Follow U.S. Lead | Fragmentation's New Discipline: Liquidity Management | Asia Comes Around to Algorithms | Low Latency Not a Go-It-Alone Proposition
Low Latency Not a Go-It-Alone Proposition
June 18, 2007
In response, providers are rearchitecting their solutions and seeking partnerships to enable light-speed processing end-to-end.
The accelerating need for speed stems from the fact that automated, algorithmic applications are made to sift voluminous amounts of market data to find pricing anomalies among venues, dealers, or among different securities, and then to submit orders immediately for execution. The opportunity to profit from such arbitraging may last mere milliseconds (thousandths of a second). And immediately after a trader enters a buy or sell order, its fulfillment, or execution, is essentially out of the trader's control. It becomes dependent on the network, the systems of the executing broker-dealer (or dealers), and the trading venue (or venues) tapped to handle the requests and complete the trade.
Latency--or delays in digital transmission of trade data--anywhere along the line is the enemy. But no single application or provider can eliminate it. Slowdowns can occur in numerous places in the trade cycle, such as on the Internet protocol network that counterparties may be using; trunk lines that connect data centers; or servers and applications supporting the pricing engines and market data platforms that broker-dealers, electronic communications networks (ECNs) and exchanges use to service customer requests.
Many technology suppliers now say that their systems offer microsecond trade execution times, but again, caveat emptor: They are only talking about the step they are responsible for, and not the entire gamut of linkages and systems required to complete a transaction.
Until recently, millisecond measurements were said to be sufficiently fast--especially when compared to the seeming eternity of seconds-long latency that was considered normal several years ago. But desire to win the race to increasingly short windows of opportunity has driven demand for much faster solutions. This has led to discussions among customers and vendors about partnering further to improve the whole infrastructure.
Faster Tuning
"When you get into microseconds, to eliminate the next bit of
latency, then you're talking about tuning servers and applications.
That's where a lot of the dialogue is now in the industry and with
our customers," said Mark Akass, chief technology officer of BT
Global Financial Services, a BT Group subsidiary that has three
arms: network-connectivity provider BT Radianz, trading-room turret
supplier BT Trading Systems and BT Global Financial Solutions, an
IT outsourcing service.
Without being specific, Akass mentioned the potential need for stronger partnerships among network providers like BT Radianz and hardware and application vendors, as well as "specialty firms" that can fine-tune algorithmic applications and servers. "Having fast boxes is clearly an obvious thing to do," Akass said. "But then it's how fast the storage works, and how well the algorithms are drafted. You have to look at this end-to-end."
He said this shared, industry interest has become "a big item of discussions at the moment. We've explored [partnerships with hardware providers] and specialist providers of IT tuning services. They may or may not happen."
