The Gateway to Algorithmic and Automated Trading

CME Group: From Mainframe to HFT

Published in Automated Trader Magazine Issue 26 Q3 2012

In nearly thirty years at CME Group, COO Bryan Durkin has seen numerous technological changes. In this issue's Exchange Views, Bryan tracks some of those changes in the context of the exchange's expansion and shifts in the trading landscape.

Bryan Durkin

Bryan, you've obviously experienced, and in some cases been responsible for, changes at CME Group. Which do you regard as the most significant of these?

While there's clearly been substantial technological change since I've been with the exchange, I would say that in many respects the last five years have probably been the most remarkable. In that time we have grown into a truly global organisation and our technology has had to keep pace with that growth. A key part of that has been providing ease of access to users in more than 150 countries around the world.

At the same time we've had to extend our support to a broader and more diverse set of asset classes. Within the last five years we have brought together some of the largest trading institutions in the world by combining CME Group, CBOT and NYMEX, with most of that process taking place within barely two years. So CME Group's existing base in interest rates, agricultural products and equity indices had to be integrated with energy and metals from NYMEX, together with the other agricultural products and interest rate products of the CBOT.

All of these individual product groups have their own nuances that have to be allowed for in the associated technology used to deliver them to users. That meant that the onus was on us to ensure that the technology in Globex, our gateway infrastructure and our telecoms were kept ahead of the curve and were also sufficiently flexible to accommodate these nuances. The same also applied to functionality in areas such as risk management and credit control.

How did you handle the migration in technology terms? What were the challenges specific to the exchanges you incorporated?

NYMEX and the CBOT were different in terms of the scale of the work that needed to be accomplished. In the case of the CBOT, it had previously been using the CME Group to provide clearing services. Therefore, our focus was on transitioning all of the CBOT's products onto the GLOBEX Electronic Trading Platform. In the case of NYMEX, the task was slightly different in that the exchange was already on Globex, but was on the NYMEX Clearing system which needed to be migrated to CME Clearing.

What sort of changes in terms of hardware and other technological infrastructure have you made?

We were previously using Tandem NonStop mainframe technology to support our trade matching and mainframes for clearing, but have now largely completed a migration from the legacy platforms to distributed computing with Linux running on commodity hardware. We are now very much a Red Hat Enterprise Linux operation for our trade matching and clearing systems. With regard to databases, we are mostly Oracle, while for programming languages, we rely heavily on Java.

How has that shift away from mainframe changed things in areas such as implementation flexibility?

For the better and in a big way. Moving from mainframe has been extremely positive in terms of overall infrastructure requirements, cost of support - in terms of both primary and backup systems - and especially with regard to performance and ease of system scalability.

However, while having that flexibility is clearly valuable, we do everything possible to avoid doing things on the fly just because we can. We have a major focus on continuously monitoring system utilisation and using the information derived from that to ensure that we always have ample spare capacity in place. That applies across multiple areas including order processing, networking and connectivity (such as optimally balancing users across gateways). As a result, we're always looking to make any necessary upgrades or enhancements well in advance of actual need, so we're not in the position of having to make a sudden change overnight.

Figure 1

Figure 1

What about handling growing message volumes?

We've clearly seen a major increase in our order message volumes across all our asset classes, partly due to our global expansion and partly due to organic growth. A significant portion of this organic growth can be attributed to the rise in the popularity of automated trading, including high frequency. All this increased automation translates into additional messaging volumes that require further capacity. As a result, we devote a lot of effort to monitoring message volume levels and adjusting capacity so as to minimise latency (as well as its variance) in order to deliver a consistent user experience with regard to message throughput and order turnaround times.

While we don't differentiate between manual and automated traders in terms of the types of system access, we do require that automated trading systems are registered with the exchange. We have an order messaging policy that reflects our expectations of how users accessing our systems and trading our products do so, but this policy is not directed at any particular segment of user. The policy sets a volume to message ratio for each of our products based on the overall market, which makes clear our expectations in terms of efficient use of the system and bandwidth. We work closely with end users to adjust their order entry procedures when we see situations where they are exceeding these thresholds. Ultimately we may take certain actions, one of which is a monetary penalty.

How do you handle technology planning?

We have a dedicated architectural team that reports to our CIO Kevin Kometer. Their job is to analyse current and evolving technologies in the context of future development - typically over a two to five year horizon. They are therefore always looking for the best information on what is emerging in the technology space and how that might enhance our existing market products, business line offerings and associated infrastructure. That could range from technology that we might use to build part of new infrastructure (such as the GPU technology we are currently examining) all the way to discrete vendor products.

What's the exchange's policy as regards building versus buying technology?

We certainly build a lot of our technology internally, because we have the resources and skill set to do so. However, we are agnostic and have an open mind with regard to the technology available externally and how it might augment what we are doing, or accelerate our progress and effectiveness in reaching our goals. For example, we wouldn't be looking to write our own custom networking protocols from scratch, but will instead prefer to use generic networking technologies - such as 10 GbE switches, which we currently use for our co-location trading connections. We generally look to buy enabling tools that complement our core competencies such as middleware, database, and grid software.

Now the inevitable question - how do you feel the exchange compares with its peers as regards latency?

Based upon empirical evidence in the form of feedback from our user base, I would say we compare well. Our customers have been clear that it's not strictly about just offering the lowest latency, but also offering a predictable and reliable matching service with the highest levels of trust and integrity. The chart above (see Figure 1) shows the progress we have made with regards to latency in recent years, and we are very excited for our upcoming CME Globex release that will further increase the predictability and performance of the matching platform.

What is CME Group's position on the CCP clearing of OTC products?

We are very explicit that our focus is within the IRS, OTC Clearing and CDS markets. We started with a good base here courtesy of the acquisition of NYMEX and its ClearPort capability, which has historically been used for clearing OTC energy transactions via a central counterparty. We have now enhanced this offering to additionally incorporate interest rate swaps, credit default swaps, foreign currency, OTC metals, agricultural and energy - and the list will continue to grow. Obviously, the ability to integrate clearing on exchange traded and OTC products has major benefits in terms of the capital efficiency associated with margin management.

Finally, what does CME Group offer in terms of co-location services?

We undertook a significant build out about 18 months back and went live with a co-location offering earlier this year. This enables any of our current or future users to place their technology systems in the same data centre as our trade matching systems. Our intent from the outset has been on offering all users the same level of access and performance, so there is just the single offering that provides identical wire speed with equidistant wire length and round trip times to all.