The Mainframe Innovation Drag
I spend a lot of time around innovation these days, both real and perceived. I have to look at a lot of companies advertising innovation on behalf of clients for both investment and usage opportunities. If I look at the innovation curve based on the difficulty of the problems people are trying to solve in our industry it wouldn’t be very steep. User interfaces, microseconds, correlations. These are not hard problems with modern technology. Want to solve a hard problem? Eliminate the mainframe.
The mainframe is the innovation equivalent of that downhill headwind on your bicycle that just sucks all the joy from the prior climb. You think you’ve earned a reward to fly a little bit and instead you’re pushing, pushing, pushing.
I spent a few years running front-office product at Comprehensive Software Systems, which is now called Talisys. Talisys was funded by a consortium of eight financial firms wanting a modern clearing system to get rid of the batch processes, 3,270 screens, and technology only slightly younger than punch cards. The idea was sound but having been on the inside, answering to a half-dozen masters with competing priorities was painful.
Fast forward to 2014. There are many individual firms with their own innovation funds, councils, and advisors. The only way to push through that headwind is going to be coalescing a group of motivated people around the objective of eliminating the mainframe. Riders who group together are called a peloton and cyclists who are part of that group expend 40% less energy than they would riding alone. Mainframes are the systems that belong to operations groups. And operations groups, well, not exactly the first place a business wants to spend money.
It might be time to try that consortium experiment again only with a vastly more modern approach. I wonder how long it would take to build an amazing back-office system if every self-clearing firm decided they were willing to participate in a cloud-based open source project. I also wonder what would happen to T+3. There’s no way some engineer is going to looks at a stack with real-time processing and say, “Now how can I build three days of latency into this?”
I can tell you some of the mistakes I learned from my consortium experience that would have to be no-no’s worthy of capital punishment:
1. The goal is to eliminate the mainframe, not to build everything or be all things to all people. We aren’t going to build a FIX engine when they make a little thing called QuickFix. Stuff like that.
2. We don’t care if the current participants don’t trade a particular asset class or market today. We are not going to build a system that will preclude participants from doing that in the future. If we really want to be crazy, we’ll also assume they don’t speak English.
3. You will pay for commercial security audits on a regular basis to ensure these do not become high maintenance one-off events during each deployment.
4. We are not going to create vendor lock just because current participants all use the same market data, middleware, or anything else.
If you want to see genuine innovation take off in capital markets, free the big firms from their mainframes. Unlock their data. Give it to them faster and cheaper. Give them an easy way to extend it and learn from it. Then let the real innovation begin.
Adam Honoré is a CEO at MarketsTech, LLC. Prior to joining MarketsTech, Mr. Honoré was the managing director at NASDAQ OMX responsible for global business development and the ISV program for FinQloud, and research director of the institutional securities ... View Full Bio
Tweet This