A simplified view of this customer’s data processing is that there are two halves: the front is the trading platform that supports the trade floor and computer trading. The back is known as the “post-trade” processing platform. The back half does everything from settling trades to market surveillance to interfacing with all the exchange members and the clearing houses. The post-trade platform is where the heavy computation happens, and this customer is reimplementing all of it in Ab Initio.
The legacy post-trade system dates back 20 years and consists of more than a million lines of Cobol code. An increasingly outdated behemoth, it is not only difficult to maintain and enhance, but, complicating matters, the original coders are retiring. This means that every change to this platform is a big deal, and naturally the business people are frustrated that they can’t get the functionality they want.
In an effort to lower its costs, the exchange decided to move the post-trade system to less expensive Unix servers. Their first try was to use Unix rehosting software for Cobol. In principle, this should have been easy: just take the Cobol code and relocate it to the Unix servers, where it would run with no changes. This would not modernize the code, but it would achieve a key goal of getting off the mainframe and onto much cheaper Unix servers.
But even after more than a year and a large expenditure of effort, the results were not pretty. Too much code required adjusting, the adjustments required significant testing, and in the end the resulting code didn’t run fast enough. The environment was extremely complex and hard to manage, and there were reliability issues. Plus, the output was still Cobol, so none of the other business goals were addressed.
The good news is the exchange discovered Ab Initio. Ab Initio has technology that will automatically translate 70% to 90% of a large Cobol application to Ab Initio’s graphically driven Co>Operating System. Instead of code that is hard to read and maintain, the entire million-plus lines of Cobol were translated into easy-to-understand graphical images and business rules. Because of the general-purpose nature of the Co>Operating System, not one piece of the resulting system required external coding, shell scripts, or stored procedures.
The entire effort of converting the million-plus lines of Cobol to Ab Initio took just 10 months from start to finish. The sections of the code that could not be automatically translated (mainframe data structures like VSAM that don’t exist on Unix, and arbitrary GOTO statements that create infamous “spaghetti” code) were handed off to a small offshore team. The total end-to-end productivity was more than 1,000 lines of Cobol per person per day to get from the original code to QA.
Another critical requirement was interfacing with the trading message bus. This bus can burst at rates of hundreds of thousands of messages per second. Needless to say, every message has to get through, and in the proper order. To make this possible, the customer had built an elaborate infrastructure in Cobol. As a result, the customer’s IT professionals were concerned about moving forward. Having been burned before by software companies with exaggerated marketing claims, they were naturally skeptical that Ab Initio could not only translate and run millions of lines of Cobol code but also process data at rates of hundreds of thousands of messages per second.
But Ab Initio does not exaggerate. It doesn’t have to. Not only was the customer’s complicated and unwieldy code successfully translated and put into production in just 10 months, the Ab Initio applications were benchmarked at many hundreds of thousands of messages per second. And this was easy because of Ab Initio’s built-in robustness.
Now that the post-trade platform has been rehosted to Ab Initio, this customer is able to keep its costs lower while still responding to the needs of the business far more aggressively than before. Plus, it can do so in full confidence that its systems will scale in a robust manner to whatever level is needed.
From this, even more ambitious projects are under way.