Technology report: Low latency systems aim to meet HFT demand

Copying and distributing are prohibited without permission of the publisher

06 October 2011

High frequency trading (HFT) firms, trading a large volume of contracts on a small margin, can find one small price movement potentially bank-breaking. That makes the choice of technology, which turns a firm’s orders into fills, crucial. But in a crowded marketplace, how do you decide what is right for you? Dan Barnes, looks at the options.

Low latency means high speed; firms’ algorithmic derivatives trading strategies rely upon split-second timing to work, with orders communicated as reliably and with as little resistance as possible.

Keeping up with the Joneses

Two research papers released by TABB analysts this summer, ‘High frequency swaps trading: market-making and arbitrage’ by Kevin McPartland, and ‘EU equity options market structure; opening the door to high frequency flow’ by Will Rhode, argue that new regulation in the form of Dodd-Frank, MiFID II and the European markets infrastructure regulation (EMIR) is likely to boost trading volumes in both the US and Europe as a wider range of contracts become accessible to trade on electronic markets, a necessity for HFT.

For firms looking to exploit this opportunity, or simply to protect themselves against increased price volatility, low-latency connections between markets, data providers and counterparties are needed to ensure that trading decisions are executed correctly.

There are a range of technology providers who offer trade gateways, the systems that enable the trading and the market connectivity necessary to facilitate high frequency trading. It must be stressed that to achieve truly low latency, from order to fill, requires optimisation from the point of an order’s origin, e.g. the algorithm reacting to a price movement, through the hardware that delivers the message from the desk to the market, the processing within the market, and back.

Not all of that is within the trader’s control, but if he is careful in his selection of market and telecom provider, he can use the trading gateway effectively.

Need for speed

“The roundtrip latency for our systems can be a hard to measure because part of it depends on the trading logic; sometimes we write it, sometimes our client writes it,” says Jesper Alfredsson VP of product management at gateway supplier Orc Group.

With that caveat, he estimates that for a market making HFT firm, using an Orc trade gateway, the latency would be between 20-50 microseconds; for a firm running an arbitrage strategy it would be closer to 10-30 microseconds.

“HFT arbitrage trading is the most extreme case for low latency where you might be trading a specific future against another; for options market making you don’t need as low a latency,” he says

Orc Group offers four modules; Orc Liquidator, Orc Market Maker, Orc Spreader and Orc Trader. Liquidator is a server-based platform that can be used for trading with, developing and deploying custom automated trading strategies, and is capable of running hundreds of complex strategies. It is used by market makers, high frequency traders and volatility traders at proprietary trading firms, hedge funds and investment banks.

Market Maker is a server-based market-making product, which allows multiple quoting engines to be monitored from one screen, and offers automated hedging and market taking functionality. Orc Spreader is a low-latency, used specifically by futures spreaders to provide trading logic to automate trading. Orc Trader is the firm’s front-end, which itself enables customisation of pricing, quoting and volatility management.

Across the spectrum

Dan Smalley, director of business development, enterprise, at systems provider Fidessa says that his firm, which does not provide off-the-shelf solutions, offers instead components that can be used to develop incredibly high-speed systems, but that there is a trade-off between latency and resilience.

“At one end of the spectrum we could provide a co-located platform, with limited functionality, designed on a bespoke basis with risk checks that operate in a microsecond range,” he says. “It is difficult to compare like-for-like as there are so many variables involved but that could be a sub-50 microsecond roundtrip.”

“At the other end a broker with a number of HFT customers would not just want to be fast, it would have very specific requirements from its clients, to be consistent, to not have any outliers, to include specific risk checks,” he explains.

The Fidessa model comprises a lightweight front-end, which allows visualisation of data from servers around the world, a sophisticated order management layer, which deals with the complex routing of orders across multiple exchanges, users and hubs, as well as programme trading, a middle-office component and an algorithmic framework.

SunGard’s trading package is built around the Valdi brand, which encompasses the former GL Stream suite for algorithmic trading, acquired with software provider GL Trade in 2009, and the SunGard Global network that offers market connectivity.

Its core business supports electronic brokerages and their clients rather than HFT firms specifically, offering trading gateways, trading workstations, risk management, smart order routing and algorithmic trading software all available in the suite.

“Like most market access providers until relatively recently, most of our software would result in roundtrip times in the low numbers of milliseconds,” says David Morgan, marketing director, trading and client connectivity at SunGard. “We’ve done a lot of work in the last year on the internal architecture of the network and the software, in some case removing processes entirely, in others joining processes up where they previously ran across two systems adding latency. We’ve tuned the systems so that roundtrip times are now down to hundreds of microseconds.”

Custom strategies

Trading Technologies offers systems to the full range of traders from big brokers to retail traders. Its X_Trader flagship product enables a trader to operate on multiple markets from a single screen and the Autotrader and Autospreader components allow a user to implement new strategies, or create and execute synthetic spreads, without having to use complicated programming.

Primarily it deals with futures and some equity options, and connects via its managed hosting solution, TTNet, to data centres in Tokyo, Singapore, Sydney, Chicago, New York, London, Frankfurt and soon Sao Paolo, to allow collocated services.

“We’re focussed on high performance, high frequency trading technologies, but at scale; frankly it’s easy to offer low latency HFT for one person trading one strategy,” says Jeff Mezger, product manager at Trading Technologies, who says roundtrip latency for the trades is in the double microsecond range.

The firm has recently begun to focus on server side automated trading, where for example a trader in London can access a co-located server based in the Sao Paulo data centre via TTNet so his strategies are traded at a latency equal to that of a local Brazilian firm.

“We’re about to launch ADL, which is a visual language that will allow a trader to develop complex algorithms and launch them to an execution server sitting next to the exchange’s matching engine,” says Mezger.

Many of the system suppliers focus on customisation reflects the commoditisation of speed technologies says Mark Fischer VP of product management at CQG, provider of trading front-ends CQG Integrated Client and CQG Trader which use a network of hosted exchange gateways to connect to over 40 derivative exchanges.

“Every service provider has to be fast in order to compete in the market place,” he says. “However, the law of diminishing returns is beginning to apply to speed. The cost of getting the next millisecond of latency out of the system is going to start to be too expensive.”

Orc Group’s Alfredsson concurs. “There has been a race to zero; now we are seeing some firms pulling out the speed race because it is too hard to make money,” he says. “Even if you are fast, you won’t get every trade, so you need scale to work across lots of instruments and that’s more for the big guys.”

“Traders are now turning to smart trading systems,” adds Fischer. “The strength of their analytic tools and their ability to quickly build sophisticated trade systems is becoming as important, if not more important, than speed to the market.”

Highlights from the October issue of FOW:

News

News analysis: Data gap remains in commodity speculation row