CA’s App Development “Wind Tunnel”
Testing software components on a simulated network is the only way to deliver good software, says CA’s Shridhar Mittal
Business software developers face a major challenge – making sure their software will work on the company’s networks and systems before the application is delivered.
New applications have to be delivered quicker than ever before and are used by more people through the cloud. But developers just have to deliver them, more or less trusting they will work, unless they have a way to properly test the thing first.
That’s what Shridhar Mittal hopes to offer developers, with a concept called “service virtualisation” that simulates a company’s IT networks giving developers a place to test applications, so they can develop and deliver faster. Mittal was CEO of ITKO, a company which has pushed the concept hard and was bought last year by CA last June
Wind tunnel versus Heathrow Airport
ITKO was formed in 1999. It started out with the idea of automated testing, but quickly realised that the real difficulty was giving developers a platform to test on, since no firm would risk crashing their live networks with an application still under development.
Mittal arrived as CEO in 2005, and since the company was bought has been general manager for service virtualisation, maintaining the company’s brand identity, and that of its product LISA, inside the giant CA corporate machine.
In fact, ITKO is something of a rising star in CA. It’s a central part of its strategy for migrating applications to the cloud. At the CA Expo event in London, where TechWeekEurope met Mittal, visitors we spoke to were genuinely interested in the concept. One LISA user, the Nationwide bank, had a session in the customer forum.
Service virtualisation is so new Mittal says he still often has to explain what it is. He is often asked the difference between service virtualisation and hardware virtualisation provided by hypervisors such as VMware. People say “we don’t need service virtualisation, we already have virtualisation,” he tells us.
Is that really true, we ask? Surely the difference is obvious! Groping for a simile, we suggest the difference between service virtualisation and hardware virtualisation is like the difference between a flight simulator and Heathrow Airport. Or maybe the difference between a wind tunnel and Heathrow.
Mittal agrees: “It’s apples and oranges,” he says and picks up the wind tunnel analogy. “You don’t test a plane by building it and flying it, you put the wing in a wind tunnel.”
The wind tunnel tests the way the wing responds to wind, but it also tests how it attaches to the body of the plane and the effects of loading and stresses: “We can simulate each component and test how the system will work when it is put together.”
This testing of components may be a big strength as it can eliminate a lot of wasted effort: “If a component doesn’t pass that criterion, there is no point in sending it on to the next stage.”
The system gives out data in the same form as the users’ normal performance monitoring applications, so reports can be generated and compared with running software, says Mittal. It can provide a model of the user experience, but it doesn’t always have to because, like the wind tunnel, it tests the software one component at a time.
Service virtualisation is a patented concept, Mittal says, but to promote the idea, the company is running a vendor-neutral site at servicevirtualisation.com. “We want to educate the market on what service virtualisation is,” he says. The site lets people ask questions, and he hopes to discover from people’s experiences whether the idea really works as promised.
With only 150 customers for LISA, the market is still in its embryonic stage, and he expects it go grow quickly. Mostly, these new users are replacing older, “brute force-esque” methods of testing, some of which are slower and involve a lot of expensive resources, so the potential for expansion is big.
“Most environments today are taking too long to deliver on innovation,” says Mittal. “Testing and development is poor quality and it is very high cost, but the business says it has to get things out faster, with a Facebook, Google or Apple level of quality.”
New markets that will boost the need for simulation will include smart meters, travel and logistics applications and the move of public sector services onto the cloud, he says: “Where there is complexity, that is where our value proposition is. Applications have to deal with a greater underlying complexity and deliver a more simple service.”
The need for testing goes up as companies start to roll out mobile applications, he says: “The volume of interactions has gone up 10 or 20 times. When a company launches a mobile app they have to simulate networks with ten times the demand.”
They have to simulate real situations and fantasies, and especially the negative conditions, says Mittal: “What happens to my application when the back end slows down? What happens when it goes really really fast? For every one positive condition they test, they have to test nine negative conditions. And it is only in simulations that we can test positive and negative conditions.”
Other vendors are starting to offer similar products, but he reckons CA is ahead: “We are on our fourth generation, having learnt from 150 customers how to make it bulletproof. There is nobody who really does what we do today.”
Given that, he thinks LISA will “dominate” the space – but the CA public relations guy catches his eye. Mittal shrugs and moderates the term to “lead”.