Optimizing SOA

Hi, my name is Christopher Lochead from Mercury, and today we're going to talk about SOA, and specifically how we optimize for SOA. Now, as you know, there's lot of benefits to SOA. However, today's solutions can rapidly turn into tomorrow's problems. So, let's talk about the pitfalls over that, and how we avoid them.

If you think about what SOA is really about, we used to have a paradigm where people built very big chunks of software applications, kind of model after the client server apps. What's happening is applications today are fragmenting into small reusable components, so that we can get faster time to market, more interoperability, better integration and ultimately, use of the Internet. All that's a very good thing. However, it's changing the way people think about the software development life-cycle. In a traditional approach, what you did with the SDLC is, you had design up here, you had development over here, and that's where all the attention was in R&D. Then, you had delivery over here, which is all about quality and testing, and then you had management in production. In the old world, that it's kind of a joke one of our customers at Mercury shared with me is that, over here in design and development, people get drunk and over here in operations, they get the hangover.

Now, what's about to happen with SOA is the amount of drinking over here as we compononize apps is going to increase the throughput or the rate of change in terms of new pieces of functionality that get ruled out here. So, what's going to happen is today we have 30,000 changes a day in IT operations, new applications, new databases, etc., as we fragment applications, the number of new pieces, components in SOA apps that go through production are going to increase. So this rate of change is going to go up.

So the question is, how do we deal with this? What people are talking about doing is creating if you will, an optimization layer to really govern the approach to a couple of critical things and that's really all about quality. It's about performance, and it's about availability, and if we can govern those things to make sure that we engineer quality in, we do rigorous performance and quality testing in pre-production, and most importantly, we manage all of these SOA components in production against this set of business service levels, then we can make the big shift that's required which is simply stated from making stuff to making stuff work.