Financial Vertical: Latency and Jitter Really Matter

It might not be precisely the case that latency and jitter performance are more important than bandwidth or physical redundancy, but in some verticals, that is almost the case.

The financial services segment is a prime example, though content-based enterprise customers run a close second, where it comes to the features most important to them when adopting Ethernet-based bandwidth services.

Some of the reasons Ethernet gets chosen are obvious: it is a very well understood technology. It is an affordable technology, compared to some other alternatives. IT departments already have a large investment in Ethernet technologies and it is well understood. Also, quality of service now rivals those of asynchronous transfer mode or SONET transport.

“So by delivering a carrier Ethernet circuit into their premise we are giving them something that the IT users understand. Iit gives them something that they can plug in easily into their systems and then can trouble shoot very effectively,” says Nick Vermeer, Rockefeller Group Senior Network.

Ethernet pipes also can be purchased in larger sizes at lower cost, Vermeer adds. “So per bit it’s most cost effective to buy Ethernet than any of the other services out there.”

Ethernet also is far more granular than any other protocol. “We can deliver a 100 megabyte pipe to a customer that only needs three megabytes to start with and then we can grow them over time,” he says.

Still, Ethernet makes sense only because it supports the core business processes, built around speed of transactions. Not many industries are as sensitive to time delays as is the financial services business, especially those segments concerned centrally with market transactions. Almost by definition, information delayed is money lost.

In the trading world, “they are trying to trigger a transaction that is going to result in a purchase of thousands or hundreds of thousands of shares,” says Optimum Lightpath VP Glenn Calafati. “Missing the transaction by just a nano-second could cost the financial institution money, so they are always looking for a way to save the time it takes to issue the transaction so they beat their competitors to the actual market and get that transaction on the wire first.”

Low latency, in other words, is a feature directly tied to the core business process of trading, says Steve Kammerer, IPC VP.

“And that means low latency is the priority,” he notes. “Ethernet is a way that way we can provide the low latency.”

“What is important to them, as you might imagine, is ultra-low latency,” say David Strauss, Optimum Lightpath VP.

IPC also does something rather unusual. It prepositions its network so that the major trading partners already are “on-network.”

Say an important new hedge fund sets up. “Other people want to connect to that hedge fund company, so the value of our network isn’t just the services we provide. It’s who already is on our network,” Kammerer says. That means access to trading partners in a matter of hours not weeks or months.

“A trader is not trying to use our services to communicate with one of their peers in the same company; they are trying to create an opportunity for them to be more competitive,” says Kammerer. And that can lead to behaviors one doesn’t see every day.

“We have one customer for whom we can do a New York-to-Chicago Ethernet circuit with just under 40 milliseconds delay,” says Kammerer. “That firm wanted to buy up all that bandwidth on our backbone.” When IPC asked what applications could conceivably require that much bandwidth, they answered that they didn’t need it; they only needed a fraction of that bandwidth but didn’t want that bandwidth hanging out there for the competitors to get.

Algorithmic trading likely accounts for some of the new emphasis on low latency response, Kammerer says. Basically, more of the trades are executed by computers, not people.

In 2006 a third of all equity trades in Europe and North America were generated by a computer, for example.

The latency factor then becomes a huge driver of how fast a computer can identify a market anomaly and respond before the anomaly is corrected.

Traders often “waste” voice circuits as well as bandwidth. Where most business users use concentration, so there are fewer trunk lines than users, traders often do the reverse. “They have more outside private voice lines than people on the floor,” Kammerer says.

The reason is that outside lines connect important trading partners. Some lines might not have any actual trade activity on them – most of the time, in fact. “I might have a button on my phone that’s directly connected to a trading partner that I have in Boston,” says Kammerer. “I might not use that button more than once a month.”

But the point is that when that partner wants to call and make a trade, there’s no question about the availability of the circuit. A hedge fund might have 20 private lines connected to various trading partners and the broker pays for the lines, not the hedge fund, he adds. The point is that the large brokers use the connections to make it easier for the partners to place orders.

If a trading firm is important, they are going to want to connect to 100 to 200 of the top institutional broker firms and we have already got those firms on our network, Kammerer explains. In many cases, the traders don’t actually have to pay for the connections, as the major traders supply those connections on behalf of their business partners.

Financial customers are unusual in other ways. Moves, adds and changes are an area where competitive advantage can be gained. Say a firm moves its traders to a new floor in the same building and circuits go down for a week during a move.

The damage there is that the trading relationship is lost, not simply that there is an interruption of service.

Also, latency is an issue, but perhaps not as big an issue as jitter. “A little bit of latency on the pipe tends to not be the issue,” says Rockefeller’s Vermeer. Jitter causes packets to arrive unpredictably. And that is more the issue for many customers, he says.

“Jitter kills all of your real time applications,” says Vermeer.

Ultra-low latency also “lends itself extremely well to the health care segment,” says Strauss. “If you look at their needs with regards to the radiologist, for example, there’s a need to send images rapidly, with the clearest of resolution in order to make a basically life saving or emergency diagnosis.”

So health care is a natural, he says. Other key verticals that Optimum Lightpath thinks are logical extensions to low-latency Ethernet service “clearly are government agencies, school systems, professional services and specifically the legal community.”

Obviously, many of those kinds of customers require very low-jitter transport, as they make a business out of high-quality, real-time video. But legal vertical customers also increasingly are demanding real-time services.

“The legal markets have been demanding more services because a lot of their applications are going on line, so their legal searches, their filings and such need to be more reliable,” says Vermeer. The difference from the financial vertical is that legal customers require resiliency and availability more than jitter performance.

The main take-away is that it increasingly is the case that latency and jitter, as well as robust physical route diversity, are what drives buyers, not necessarily bandwidth or the price of that bandwidth. Though most verticals are not yet as demanding as the financial segment customer, that’s a big change. IP