HFT on the Cloud, Noise Filters and Traffic Monitoring
Thu, 09 Dec 2010 18:45:00 GMT
An Interview with Ken Yeadon
In this interview for the High Frequency Trading Review, Mike O’Hara talks to Ken Yeadon, Managing Partner of Thematic Capital Partners LLP, a venture capital business specialising in the trading infrastructure space and the application of web 2.0 technology, SaaS and Cloud Computing to democratising access to professional investment resources. Mr. Yeadon is also a non-Executive director of Fixnetix Ltd, a leading provider of low-latency market connectivity, data and co-location facilities, as well as serving on the board of several of Thematic’s portfolio companies.
High Frequency Trading Review: Ken, maybe we can start with you telling us about your own background and who you are.
Ken Yeadon: Well, I was basically a career HSBC employee, literally right from university. I started with them in 1985 on graduate rotation and within about three months I’d worked out that traded markets was what I was interested in. So I joined the first iteration of HSBC’s OTC interest rate derivatives trading team.
By virtue of knowing how to use a spreadsheet as much as anything else, I got very much involved in both the trading side and the modelling. We really were making it up as we went along, deconstructing things and solving all the fixed income yield curve construction and stuff like that, pretty much from first principles.
Eventually I ended up going to New York and running a cross-product arbitrage trading desk out there, doing things like medium term forward foreign exchange traded against interest-rate swaps and basis trading and reasonably technical stuff. Effectively what we were doing was algorithmic trading, where computers were involved in calculating the valuations but the actual execution process itself was orchestrated by human beings.
Then, by virtue of the fact that the Yen market was very interesting from an arbitrage perspective, I ending up knowing quite a lot about the Japanese monetary system and I got asked to move out to Tokyo to run the Bank’s interest rate trading operation there.
I was working at the time for a gentleman called Stuart Gulliver who’s since become the CEO of HSBC. He was the treasurer at HSBC Tokyo at the time and I got pulled over to effectively be his right hand man. That would have been about ’92.
Stuart got promoted and went off to run HSBC Asia Pacific based out of Hong Kong and I ended up as the country head and Treasurer in Japan before moving to Hong Kong to start up their OTC derivatives and fixed income trading operations there, which at the time where nascent.
Our approach to this business was largely based on rolling out a common operational and analytical template across each regional centre, and balancing local and international market activity. We delegated trading authority to people on the ground under common risk management, infrastructure, and processes.
I found myself gravitating more and more towards the electronic “supply chain” side of contributing and managing liquidity, and so, jointly with a colleague based in London, I sponsored the beginnings of HSBC’s electronic foreign exchange platform.
HFTR: What year was that?
Ken: That would have been around ’98,’99. That project has obviously been through a few iterations since but it’s still ongoing. The platform itself has changed but you can see the original footprints all over it.
In 2001, I cut loose on my own and started a consulting business, initially around e-commerce and electronic trading infrastructure. At the same time I started angel investing in the financial technology space. Over the past nine years I’ve changed the balance away from the consulting side and more and more onto the angel investing. So, we’ve ended up, myself and my partner, running a small venture capital / private equity business, which has two specializations. One is trading technology and the other is the application of social media and digital media technology to professional contexts, particularly in finance, although that is now developing into other verticals too.
During that nine year period of consulting and angel investing I ended up involved as a non-executive director for Currenex and, together with a few other people, started a high frequency trading business. We had a team out in Massachusetts basically market-making foreign exchange and in particular doing things like box trades and cross trades. Very low risk, very small ticket size, very high volume. Just picking up pips here and there via market making and arbitrage. That did quite well for a while but then when the beginnings of the credit crunch started, access to prime brokerage and leverage started to dry up a little bit.
However, we discovered that what we’d built was an extremely effective infrastructure for the operation of connectivity, low latency trading, strategy-backed testing, strategy operations and risk management. So that eventually turned into a joint venture with ICAP which is still ongoing. There’s a company called Trading Cross Connects, which is an incubation business for high frequency funds, operated by ICAP, which has taken that core infrastructure and developed it further into a turnkey platform for funds.
We’re also one of the investors behind Fixnetix, a leading provider of connectivity, co-location and trading services, and we’re investors behind a company called Smart-Trade, who offer a very sophisticated and flexible liquidity management platform in use by several top-tier banks. Those are the two most directly relevant ones, but also just participating in this space means you explore some fairly arcane specialist areas of technology, which aren’t necessarily directly related to the high frequency trading space or trading infrastructure, but what you appreciate by virtue of that background is how they can be applied to that space, which obviously helps you understand the market for them.
HFTR: That’s quite a career history! One thing you didn’t mention however was the subject of cloud computing, which I understand is also something that you look at quite closely.
Ken: That’s right.
HFTR: So I’d be interested to hear your thoughts on the intersection between the high frequency trading space and the cloud computing space.
Ken: Cloud computing is literally just a deployment mechanism. So, compared to having your own servers in your own data centre, you’re basically renting virtualized space on somebody else’s servers in their data centre. That doesn’t tell you anything about where that data centre is, how its optimized, or what types of activity it is designed to support, so when you look at the popular mainstream manifestations of cloud computing, first of all they’re very large, they’re extremely robust, and they are really pushing the envelope in terms of data centre management and “just-in-time” system availability. Those guys have led the charge in terms of the administrative interfaces and the security policies and all of the operational aspects of running data centres in a much, much more efficient way. Those are all issues pertinent to anyone using server capacity.
The extent to which those techniques can be applied to high frequency trading is really just a function of which data centre they’re applied to, by who and in what context. So, you could have the cloud model working in a data centre that’s co-located with an exchange, for example. There are already initiatives in progress designed to achieve exactly that.
Then there’s the types of computational problems you encounter, some are in the trade process, they’re part of the execution state model, so they need to be close to the exchange and highly optimized- generally they don’t lend themselves (yet!) to a cloud-based approach, but there’s an awful lot of data mining, analysis, back testing, the preamble to strategy generation where you’re basically using tick level historic data, where there’s absolutely no benefit to being co-located to the exchange. What you need there is access to massive amounts of computational power in an unpredictable way, on demand, at reasonable cost, and for operations that may or may not turn out to be productive. Flexibility is the optimum, not necessarily raw performance.
For example, if you’re back testing tick level trading strategies, there may be a period of eight or nine hours when you need absolutely shedloads of computing power, shedloads of storage, shedloads of bandwidth, but basically because you don’t do that constantly, the requirement switches on and off. So the economics of cloud computing for those types of analytical, offline tasks can be extremely attractive. That’s different to saying you would choose to use it in production strategies.
The biggest overhead with high frequency strategies isn’t actually the infrastructure, it’s the strategy generation. The infrastructure is becoming increasingly available as managed services at accessible cost, but even if you’ve built your own, once you’ve done it, you’ve done it.
Strategy generation on the other hand, constantly pushes the envelopes because you don’t put something into a full production implementation until you’ve validated it, until you’ve established that it’s got some robustness and that it actually works as a trading strategy. Agility is always going to be a major part of the equation.
So, it’s a very experimental process. You’re trying all sorts of things, often for protracted periods of time, some of which work, some of which don’t. You don’t want to be acquiring fixed costs. You don’t want to be having to invest in massive data infrastructure for something that may or may not work, where the majority of your attempts may get discarded. It’s very Darwinian, so what you need is a great deal of flexibility and agility in your access to computing power. I think cloud computing lends itself to that very well. I think the approach will gradually be applied more broadly, and expect to see specialised versions of cloud offerings designed exactly for trading.
HFTR: So how do you see cloud computing services actually being implemented in the HFT space?
If you are not a registered user, click here to register for free