The Gateway to Algorithmic and Automated Trading

A Greener Shade of Blue

Published in Automated Trader Magazine Issue 21 Q2 2011

Recent events at Turquoise, such as its acquisition of derivatives platform EDX, certainly appear to have piqued plenty of buyside interest. In fact, to judge by the number of times its name keeps cropping up in conversations with Automated Trader’s readers, you could be forgiven for assuming that rainbows were now monochrome. So we thought we’d better follow up by packing off Automated Trader’s Founder, Andy Webb, to talk to Adrian Farnham, CEO of Turquoise.

Adrian Farnham

Adrian Farnham

Adrian, I'd like to start by asking about Turquoise's customer demographic. Is there any particular category of trading participant (or combination of categories) that you are keen to encourage?

I think it's a combination and not necessarily a static one either. In the context of equities we are a pan-European platform, so we appeal to firms with pan-European volumes or strategies. Our low latency lit market is relevant to firms that have aggressive strategies (liquidity takers) and are looking to achieve high fill rates when taking prices on the screen. However, that low latency environment is also appealing for those firms that are looking to post prices and need to be able to adjust their quotes quickly in order to manage their risk effectively. For our Midpoint Book, the focus has always been (and continues to be) on agency flow from the large broker community where we have our origins.

Turquoise's relationship with EuroCCP means that we cover a broad range of stocks, and not just blue chips. For example if you look at our market share in European Mid cap stocks we have a significant presence and we are also keen to add more Scandinavian and continental-European participants.

As regards derivatives, the launch of Turquoise Derivatives and its recent acquisition of EDX has given us a strong franchise in Russian and Norwegian products. While EDX has historically been primarily a wholesale and institutional market, it is now receiving increasing focus from proprietary trading firms and we are keen to combine their participation with the existing institutional customers on Turquoise Derivatives. Our expansion into products based on FTSE and other European indices will further expand that customer base.

The interesting opportunity here is that Turquoise already has established relationships with high-frequency trading firms that EDX has not traditionally had or needed because it has effectively been a sort of niche block trading type of business. In the future we see it developing into a high-volume on-screen market, which means that there is a lot of natural synergy and opportunity there between the two categories of participant.

How do you draw the distinction in trading terms between the LSE and Turquoise? Who should use which and why?

We encourage customers to use both marketplaces, though they clearly have differing characteristics. Both markets are available via common connectivity and near-identical APIs, so once a firm is connected to one, adding the other couldn't be easier. Of course Turquoise's equity offering is pan-European, whilst the LSE's franchise is in UK and its International Order Book (Russian and other international depository receipts), so the offerings are more complementary than they are directly competitive. Where our offerings do overlap, firms choose where to add or remove liquidity based on how a combination of execution certainty, available liquidity and costs affect their own strategies or ability to deliver best execution.

The LSE has a greater diversity of participants, operates the opening and closing auctions, and hence has an attractive liquidity profile for certain strategies to interact with. Turquoise benefits from having the support of its (investment bank) shareholders, a maker-taker tariff structure, and its broad stock universe - also creating a distinct and attractive liquidity profile.

How will Turquoise handle sponsored access in terms of risk management?

Now that Turquoise and the LSE are using the same core platform this is one of the new products we shall be introducing jointly; Millennium IT has just developed a sponsored access model for exactly this purpose. (We obviously won't be offering naked access because it's not permitted in Europe and is not likely to be something we would offer in any case.) The objective with our sponsored access offering is to have an agreed set of pre-trade risk management functionality. This will ensure that all orders must pass pre-trade checks that validate order price and size, that there is a control for overall exposure and that the sponsoring firm has the ability to intervene at any time.

Clearly you would have to be a latency sensitive customer to want to use sponsored access in the first place (otherwise you could just use direct market access instead). Therefore the challenge for us is to be able to devise those controls so they have the lowest possible impact on latency. We think we can achieve this and anticipate round-trip latency including pre-trade checks being in the order of 100 µs.

Do you regard the provision of co-location services as a core business for Turquoise?

Yes we do, and we have also seen co-location customer numbers increasing. Recent events have also made it an attractive proposition for potential customers in that we now have a common network, with the Turquoise platform hosted in the LSE data centre alongside the LSE cash market. Therefore there is no marginal cost over existing LSE co-located connections for connecting to Turquoise. With all the old hardware for SETS soon no longer being required in the LSE data centre, further space will be available and the fact that the LSE, Turquoise and Turquoise derivatives are all now available in a single central London location makes for an attractive proposition.

However, it's important to note that co-location clearly doesn't make sense for every category of participant; much obviously depends upon the nature of your trading strategies. For example, customers who choose not to co-locate can still achieve a hit rate on immediate or cancel orders of 90% to 95%.

Do you perceive an increasing demand for cross asset trading and was that a factor in your decision to make the EDX acquisition? Or did you simply see it as a good book of business that you could grow anyway just in the context of derivatives?

We see opportunities for both. The equity derivatives business on its own is a significant opportunity in an area that is ripe for the introduction of competition. Other factors assisting this include regulation and exchange consolidation. The adoption of algorithmic trading and smart order routing by derivatives participants creates an environment that is able to support competition; we therefore see the sort of choice that has emerged in a fragmented equity market also becoming available in equity derivatives and other forms of exchange traded derivatives. The technology is there, the regulatory environment is there and the desire from participants for alternatives is self-evident.

Trading cost is a major factor in this opportunity. For example, if you look at index derivatives such as the FTSE or the STOXX, they are somewhere between five and ten times more expensive than non-index products, which is clearly a situation crying out for competition. Therefore purely on a standalone derivatives basis there is a strong argument for people to be looking for an alternative.

However, when you combine equity derivatives with the underlying then the situation becomes even more interesting. When you consider that the LSE data centre now covers the LSE, IOB, Turquoise cash and derivatives under one roof, then it would logically seem an attractive venue for pricing and risk managing derivatives. Therefore, in addition to the pricing benefits, there is a genuine trading optimisation benefit from having the underlying and the derivative in the same physical location.

In the light of what you've just said who do you regard as your greatest competitors? Is it the derivatives exchanges per se? Or the stock exchanges/alternative trading venues for cash markets? Or both?

We have a vision of a marketplace in equity derivatives that is able to sustain choice, which leads to innovation and service but also benefits in terms of tariff and market model. We certainly don't think that we would be the only firm operating in such an environment (we are obviously aware that firms such as Chi-X/BATS are considering an equity derivatives platform in Europe) but we think such an environment would be healthy and also play to our strengths. In the light of that, we would welcome the concept of a community of competitive venues that wished to change the status quo by being open to competition.

In addition to Eurex and LIFFE, the emergence of at least one other credible alternative venue seems almost inevitable. We would certainly relish the opportunity to compete with them head-to-head on key products and through that type of direct competition provide the sort of choice that leads to innovation and further competition.

In addition to price competition on established exchange traded derivatives, there also seems to be plenty of scope for completely new derivatives contracts in Europe. We would certainly hope to be able to innovate in that space, as I assume would any emerging competitor.

Many Automated Trader readers continually monitor the frictional costs of trading so that as these continue to decline they can deploy new strategies that current cost levels render non-viable. In which case, do you believe that Turquoise will be getting a bigger slice of the existing pie or the same share of a larger pie? And how do you think MiFID2 might affect future pie growth?

We certainly have plenty of feedback from proprietary trading firms on the implications of reducing frictional trading costs for new strategy deployment. However, in view of the financial environment post-Lehman it's hard to point to strong empirical evidence that the reducing cost of trading in Europe has driven a huge increase in volume. On the other hand, you could argue that we've been in a rather extreme phase of diminishing asset values so perhaps trading volumes would actually have declined dramatically had it not been for the reduction in trading costs!

There is certainly scope for organic growth. If you consider our first derivative product, the FTSE 100 future, the market for this is actually relatively underdeveloped. There are a variety of reasons for this as regards the UK market, with the popularity of alternative instruments such as contracts for difference or spread betting clearly a factor, but there is little question that the FTSE index contract definitely has the potential to be significantly larger. In view of
that, it definitely isn't just a case of us having a greater share of pie, but for the pie itself to be appreciably larger.

As regards equity derivatives in general, if you look at the US market for equity options versus the European market for equity options, the European market is significantly underdeveloped. Again we would hope to be part of the process that changes that and grows significant volume in equity futures and options in Europe as a result of reducing frictional costs and providing better/faster trading facilities.

As regards the equity cash markets, I think it's a bit early to tell what MiFID2 will bring, but we certainly feel the first incarnation of MiFID was a good thing. There were genuine benefits in terms of reducing spreads and costs, while the level of innovation has increased with mechanisms such as public dark pools that offer price improvement. MiFID has delivered what was intended in that customers have greater choice and there is innovation in terms of tariff, market model and customer service. We would certainly hope that nothing in the next incarnation of MiFID would in any way make these positive developments in Europe less likely in the future.

I'd now like to move from the generic to the specific and to talk a little about Turquoise's technology. One area that Automated Trader's readers have expressed particular interest in is the relationship with Millennium IT. Given that they too are a subsidiary of the London Stock Exchange Group, how arm's length is the relationship with them?

We certainly regard it as an arm's length relationship in that there is a contract for a software licence, support and development - much as there would be with any of their other clients - so contractually they are certainly independent vendors. Nevertheless, it is a very close working relationship and we also treat them as a partner. One important facet of this is that they are very keen to understand our business needs and not just our technical and functional requirements. The fact that they are pitching to other customers gives them a good understanding of the competitive environment. They are also very willing to engage directly with our customers when asked to talk about our platform design and its probable future direction.

It's certainly a symbiotic relationship in that we are clearly a valuable prototype client for them. This tends to apply in two distinct areas. On the one hand there is core functionality such as matching engines and gateways, where they have a continuous improvement concept in place anyway. This will apply to matters such as the further evolution of the matching engine and continuing reduction of its latency where they will advise us of the projected future path. On the other hand are business issues, such as order routing, where they will approach us with their order routing solution and suggest ways in which it could be deployed most effectively in our business.

Can we talk about your core technology; how are your matching engines configured and what hardware are you using for them?

We use clustering for our matching engines, which are based upon IBM servers running quad core Intel Xeon X5570 CPUs. Our matching infrastructure is built around the concept of grouping matching engines into what we call a partition, with each partition having its own built in redundancy.

What operating system are you running on the matching engine servers?

SUSE Linux 11.

That's interesting; the vast majority of exchanges we talk to these days seem to be running Red Hat Enterprise Linux on their matching engines. Was there any particular reason for choosing SUSE over RHEL?

We were actually using RHEL before with the earlier version of our matching software, but when Millennium IT moved us to our new version it was felt that SUSE was more appropriate to the combination of hardware and software we would be using.

What level of spare capacity headroom in terms of CPU, memory and network do you maintain?

We typically work on the basis of having 2.5 times the previous peak trading activity level in hand and monitor that closely. For example, we were recently seeing sustained increases in activity levels on our lit book and in response installed an additional matching engine partition so that we now have two for the lit book. We were still a long way below peak capacity, but for an operation like that we prefer to be pre-emptive so that we have plenty of time for testing. While we can quickly replace an individual server (over a weekend or even overnight) adding an entire new partition to the matching engine requires more planning. That is probably closer to a two-month job, but it only takes that long because we like to set the new partition up in a customer test environment so there is plenty of time for user acceptance testing.

What sort of network technology are you using?

We use InfiniBand between the components of the matching platform and then 10 Gbit Ethernet for our internal network and external market data. However, we're currently considering a pilot project to expand external network capacity by rolling out an InfiniBand market facing network.

What would be a typical round trip time for your platform?

Adrian Farnham

For our lit book, average latency is around 105 µs and for the Midpoint Book it is 101 µs. That represents the time taken for an order to move from the outside interface of the firewall to the matching engine and for the resulting order acknowledgement to reach the outside firewall interface on its return trip.

How do you deal with issues such as order message flooding?

We have inbound and outbound message throttles on our gateways that can be invoked to deal with this sort of situation, but trigger levels on these are set at such high levels they have never yet been triggered. We monitor activity on the gateways and load balance across them as necessary. As I mentioned earlier, we try to be pre-emptive with upgrades, so while we were nowhere near capacity on our previous gateways we have recently added a further set of them.

Finally, is there any particular generic technology you feel looks particularly promising for possible adoption by Turquoise?

At present, one area that Millennium IT are looking into is the possibility of employing field programmable gate arrays (FPGAs) to accelerate our FIX trading gateway, with the intention of reducing the current performance gap between the FIX and native protocol versions of the gateway.