Over the past five years, Chicago Mercantile Exchange Holdings Inc. (CME) has gone from electronically trading 137,550 futures contracts a day to trading almost 3.15 million contracts per day. During that time, the exchange also went public -- years ahead of rival New York Stock Exchange Inc., which recently announced plans for an initial public stock offering. Jim Krause, CIO for the CME, spoke recently with Computerworld about what it takes to stay ahead of the IT systems game and how market consolidation will affect future technology deployments.

The NYSE is buying a premier electronic trading platform, Archipelago. Nasdaq Stock Market Inc. is replacing its SuperMontage electronic trading platform with Instinet. How is the CME going to keep up with the other exchanges? You"re doing things to constantly improve your technology and improve your response time. One of the challenges is that in an automated environment, let"s say you reduce the response time for an order by half. Well, then the next order from an automated trading system is coming to you that much faster, and therefore, it"s like your capacity or extra capabilities are used up overnight. I think all the exchanges are experiencing that problem. I think the big issue you"ll see with automated trading -- I know Nasdaq has this issue because we"ve talked to them -- is the proliferation of market data, especially when you start trading options electronically.

One of the big things we"ve been working on with the other exchanges is developing highly efficient standards and capabilities. ... [That way], you can minimize bandwidth growth impacts and maximize the amount of information going out the door to traders so you can make that as timely as possible. That"s one of the things the industry is wrestling with -- how to control the growth of market data. Just like a standard called FIX [Financial Information Exchange] has been developed for order flow in getting orders into and out of systems, there"s going to be a standard developed for that market data as well.

With its acquisition of Archipelago Holdings Inc., the NYSE is going public. You went public more than two years ago. How did that affect operations and the use of technology for trading? That"s why I"m saying they"re all copying us. Going public and having to react to shareholders" concerns, the ability to grow the marketplace, etc. ... has put more demands on IT to deliver functionality and capabilities. I think we met those challenges, though maybe the growth rates could have been better if we were faster. It"s hard to say. I think we realize the electronic market is the key to the exchange in the future, and we"re willing to invest in the technologies to expand the trading to keep it a premier platform. ... In the end, it is still all about shareholder value.

The NYSE is looking at keeping a hybrid system that includes both the verbal "outcry" system and electronic trading. Do you think that can work? We"re still in a hybrid mode. If you look at what makes a market -- is it technology or something else? -- I"d argue it"s something else and that something else is liquidity. The ability to get in and get out of a market when you want to and when you need to. Electronics can certainly help that. What we"ve done is figured out that instead of bifurcating liquidity between open-outcry and electronic [trading], we figured out how we could marry those two into a single market. It doesn"t matter where you execute. You can still get the best price.

My wife likes going to Saks Fifth Avenue instead of Kmart. Even though the price might be cheaper at Kmart, she wants the service. Obviously, the customers have been deciding to move electronically, but there are also various things that [make trading] complex, and you may want to talk to a human being.

Are you looking at increasing processing capability? We moved over to your normal commodity, Linux-based Intel processors. I"d like to say there"s both horizontal scalability and vertical scalability that we need to address. If you trade two products or multiple products, then you can put one product on one server and another product on another server, and in theory, that"s parallel processing. That"s what I would call horizontal processing.

When you look at the fact that we probably have 15 or 20 major products when we"re trading the index, there"s a tremendous amount of transactions coming through one individual product. You have to be able to scale up. In other words, you have to be able to provide bigger and badder processors, if you will. You can"t always solve the scalability problems by scaling horizontally.

IT budgets were hit hard over the past five years. While there"s been something of an uptick recently, the purse strings are still pretty tight. What drives your spending decisions? That"s why we"ve gone to a lot of the Linux-based commodity servers. But, the horizontal scalability is not too bad. The things that become more prevalent are the software costs when you start licensing multiple, multiple, multiple versions of applications, like Oracle or OSes. That"s a cost you have to watch. Also, as servers get more capabilities and become faster and smaller, you face the infrastructure issues of power, cooling and wiring data centers. That for us is a big cost issue. Your conventional data centers, when they were built, were not built for racks and racks of 1U [1.75-in.-high] servers sitting together. The amount of heat they generate, especially some of the Intel boxes, is really where the costs are. That"s something we"d like to say we"ve had a pay-as-you-grow philosophy. As we need additional servers, that"s when we extend the infrastructure to handle that. We didn"t build this huge data center completely outfitted. We"ll step through the growth and do things on an as-needed basis.

Have you heard of IBM"s Cool Blue project and using water to cool high-end servers? Does that interest you? It"s something we would look at. We"ve known about their thinking on that for almost a year. This is going to be a problem. We have to get into that a little more. The blade-server issue that that thing is addressing is obviously a concern. We have built data centers that can handle a lot of that heat removal. We know where we were going. If you look at data centers three or four years ago, the whole idea was moving the cool air to your servers. Now it"s not only moving the cool air but removing the hot air. We"ve done a lot of work with a lot of vendors to properly maintain the cooling.

Are you considering grid computing technology? I think right now it"s really just an R&D effort. Could that technology be not only applicable but efficiently deployed in an electronic trading environment? What advantages does it have versus the straight client/server, multitiered approaches we have now? We are working with a think tank, but I don"t want to say who. But we are using some high-powered research to try to figure out what is the best approach for that going forward.

How has the use of open systems helped you reduce cost and speed development? Most of our development is done in Java. Moving it from a Unix-based platform to a Linux is a few small tweaks. The biggest benefit we got with Linux was significant reduction in response time. Initially, that had to do with the efficiency with TCP/IP stack on Linux versus what was available on Unix. Plus, the processors are faster than what was available in the Unix arena.