Financial IT Managers Ask for Microsecond Granularity
The financial industry long has been a lead adopter of the latest technology innovations, and the quest for absolute low latency communications is among the latest areas where securities firms are pushing carriers and service providers for unheard-of latency performance.
At a conference sponsored by A-Team, financial industry information technology executives pointed out that they now are looking for transaction latency performance measured in microseconds. To put that in perspective, latencies on the best optical networks are measured in milliseconds (thousandths of a second).
Asked what might be possible in the relatively near future, though not necessarily as a market-available product today, Jock Percy, Perseus Telecom CEO, suggested it might be possible to get to 14 milliseconds on the New York to Chicago route; 64 milliseconds across the Atlantic or 95 milliseconds from Los Angeles to Tokyo. Keep in mind that financial IT executives were asking for performance in microseconds, roughly three orders of magnitude more stringent than those metrics.
Even on the fastest networks, it takes seven milliseconds for data to travel between the New York and Chicago markets, one way, and 35 milliseconds between the West and East coasts. For that reason, many broker-dealers and execution-services firms are paying premiums to place their servers inside the data centers of Nasdaq and the NYSE.
Though some executives suggested much of the latency actualy was caused by applications, servers and other hardware, carrier executives said one obvious way to reduce latency in transmission links was to select the shortest routes. Some providers have an easier time of it.
John Knuff, Equinix general manager pointed out that one additional way to limit latency is to avoid placing any network gear between one firm and a trading partner. For Equinex, that means connecting trading partners within a single collocation facility.
That’s good advice even for a local, regional or international link as well, says Accedian Networks VP Scott Sumner. “Whenever possible, you want to eliminate as many active elements as possible,” he says.
But is is not often possible to connect every trading partner within a single facility. When physical collocation is not possible, RCN Metro has to figure out what locations are key to those trading partners, and make sure it has high-capacity and direct links between those locations, says Mary Stanhope, RCN Metro senior director. “Then you groom the circuits between the locations,” she says.
“After making sure you have the shortest-possible route, you also can take advantage of tweaking the dispersion compensation method, for example,” says Ernie Hoffmann, Optimum Lightpath VP.
“Network design also plays a role,” says Percy. Also, “it’s easier if the transport is at layer one instead of layer two,” says Hoffmann.
In fact, some customers now are asking for latency measurements made every second on a communications path, says Hoffmann.
Others note that measurements made as often as every tenth of a second likely will be requested. One issue financial IT staffs now are attempting to grapple with are microbursts of data that temporarily cause huge spikes in bandwidth demand.
An example is the start of a trading day, when networks are hit with a flood of orders that have been queueing up since the night before.
To analyze performance issues, network managers often have to break out a one-second period into 100 millisecond intervals, 10 millisecond intervals, or 5 microsecond intervals for investigations.
A sub-second period where a major burst of traffic occurs is known as a microburst.
Some communication executives will point out that such microburst problems can be avoided if there is enough bandwidth on a link, or even when “bursting” features are available on a connection. The issue, as always, is the balance between latency performance and bandwidth cost.
The financial industry long has been a lead adopter of the latest technology innovations, and the quest for absolute low latency communications is among the latest areas where securities firms are pushing carriers and service providers for unheard-of latency performance.
At a conference sponsored by A-Team, financial industry information technology executives pointed out that they now are looking for transaction latency performance measured in microseconds. To put that in perspective, latencies on the best optical networks are measured in milliseconds (thousandths of a second).
Asked what might be possible in the relatively near future, though not necessarily as a market-available product today, Jock Percy, Perseus Telecom CEO, suggested it might be possible to get to 14 milliseconds on the New York to Chicago route; 64 milliseconds across the Atlantic or 95 milliseconds from Los Angeles to Tokyo. Keep in mind that financial IT executives were asking for performance in microseconds, roughly three orders of magnitude more stringent than those metrics.
Even on the fastest networks, it takes seven milliseconds for data to travel between the New York and Chicago markets, one way, and 35 milliseconds between the West and East coasts. For that reason, many broker-dealers and execution-services firms are paying premiums to place their servers inside the data centers of Nasdaq and the NYSE.
Though some executives suggested much of the latency actualy was caused by applications, servers and other hardware, carrier executives said one obvious way to reduce latency in transmission links was to select the shortest routes. Some providers have an easier time of it.
John Knuff, Equinix general manager pointed out that one additional way to limit latency is to avoid placing any network gear between one firm and a trading partner. For Equinex, that means connecting trading partners within a single collocation facility.
That’s good advice even for a local, regional or international link as well, says Accedian Networks VP Scott Sumner. “Whenever possible, you want to eliminate as many active elements as possible,” he says.
But is is not often possible to connect every trading partner within a single facility. When physical collocation is not possible, RCN Metro has to figure out what locations are key to those trading partners, and make sure it has high-capacity and direct links between those locations, says Mary Stanhope, RCN Metro senior director. “Then you groom the circuits between the locations,” she says.
“After making sure you have the shortest-possible route, you also can take advantage of tweaking the dispersion compensation method, for example,” says Ernie Hoffmann, Optimum Lightpath VP.
“Network design also plays a role,” says Percy. Also, “it’s easier if the transport is at layer one instead of layer two,” says Hoffmann.
In fact, some customers now are asking for latency measurements made every second on a communications path, says Hoffmann.
Others note that measurements made as often as every tenth of a second likely will be requested. One issue financial IT staffs now are attempting to grapple with are microbursts of data that temporarily cause huge spikes in bandwidth demand.
An example is the start of a trading day, when networks are hit with a flood of orders that have been queueing up since the night before.
To analyze performance issues, network managers often have to break out a one-second period into 100 millisecond intervals, 10 millisecond intervals, or 5 microsecond intervals for investigations.
A sub-second period where a major burst of traffic occurs is known as a microburst.
Some communication executives will point out that such microburst problems can be avoided if there is enough bandwidth on a link, or even when “bursting” features are available on a connection. The issue, as always, is the balance between latency performance and bandwidth cost.
by Gary Kim
