Well, what if there was a way that you could implement changes to one subsystem and *know* that it still work with all the others, and test it under perhaps 100 times the normal load - completely independently of the rest of the system?
That's where CA subsidiary ITKO's LISA platform comes in. LISA can capture transactions passing between subsystems, and then simulate the rest of the system when a component is being tested. Sources of data include CA's Wily application performance management and NetQoS network performance management products.
This proves that the individual applications continue to work together despite the changes, and allows stress testing by increasing the rate at which the simulated load is played back.
John Michelson, ITKO's chief technology officer, told iTWire that giving software developers and testers access to LISA is like giving aircraft wing designers a wind tunnel rather than expecting them to attach a redesigned wing to a complete aircraft for testing. "It is a transformational technology," he said.
This allows the parallel development of the individual pieces of composite systems, and Mr Michelson noted that one telco had to change 38 interrelated systems before its LTE network could go live. Being able to develop and test each application separately telescopes the overall development time.
In such situations, LISA has the ability to compare the modelled behaviour with the actual behaviour when any new module goes live. The virtual services that represent the rest of the composite system to each component can then be updated without causing any disruptions.
It is also possible to model the effects of moving components to more or less powerful hardware, of changes to the performance of external systems such as credit card processing services: "the volatility of external applications confounds dev and test activities," said Mr Michelson.
For example, FedEx provides its customers with access to a LISA virtual service that they can use for testing their applications. The speed of the virtual service can be adjusted to represent the response of the real system at different times of day.
Using another analogy, Mr Michelson likened LISA to a flight simulator. Rather than trying to find a wide range of weather conditions in the real-world, pilots learn how to cope with those conditions in a simulator. Similarly, LISA can test applications under a wide range of conditions including the most arduous that are anticipated.
Other advantages include more successful integration (thanks to more thorough testing) and more effective outcomes.
"If you solve it for the biggest companies in the world, you've solved it for the little guys," he observes. It makes commercial sense to address the top of the market first, as that makes it harder for potential competitors to gain a toehold.
The product could be useful to organisations with a few hundred developers, he said, "but we don't go looking for them." Instead, the company partners with dev/test specialists who can then introduce LISA to their clients.
Existing customers in Australia include the all of the big four banks - ANZ, Commonwealth, NAB and Westpac - Telstra, and Hydro Tasmania.
No local customers were available for comment, but according to Mr Michelson big US customers have found LISA can save them millions of dollars and slash 30% from the time needed to get a product up and running. The company's customers also include AT&T, Best Buy, Citigroup, General Motors, IBM, Intel, Kaiser Permanente, Lockheed Martin, Oracle, the US Army, the US Air Force, and Vodafone.