Calls to help desk = testing, survey finds
A survey of European IT execs conducted by Compuware has shown that 52 per cent believe there is a more than even chance that applications deployments will fail when they go live.
The instant response at Vulture Central was "another statement of the blindingly obvious", which must be the result of our renowned levels of elevated cynicism, because according to Michael Allen, the company's global director of performance solutions, most people he'd talked to seemed quite shocked at the suggestion.
The particular thrust of the survey was to identify the effect not planning for performance issues as an integral part of the development process can have on applications once they hit the production environment.
"The tools are available, but most users think about performance too late into the development process," Allen said. "It is often not till they get to load testing, and many applications are put into production without testing them against the wider production environment."
In particular, this means planning for the implications of using Wide Area Networks (WANs) in the real production environment. According to Allen, many applications are developed and tested purely on internal LANs, and then hit serious load problems when they are put into the production environment, where it is increasingly common for a WAN or two to be part of the wider infrastructure mix.
"Applications that go outside the LAN and across WANs will work very differently from those that only run on the internal LAN," he said. "LAN-based applications can often be written in a very chatty way, which works fine over a LAN, but can kill an application once it hits the WAN."
This means they have to be written with that prospect firmly in mind. And the necessary changes can be quite simple. As he put it: "Rather than writing code in a 'get me an A, get me a B' style, it could be 'get me an alphabet'. This reduces the load on communications services such as WANs."
Perhaps the most telling statistic from the survey is that most IT departments (71 per cent) seem to rely on end users calling the help desks to alert them that performance problems exist. This means problems are only reported after their impact is noticed.
Given the fact that an increasing number of those end-users are now customers, business partners and others that can take their business elsewhere, this can mean serious damage being done to a company before any remedial action is taken.
One of the problems Compuware faces with pushing the performance issue to the fore is that the word itself has acquired a rather narrow definition geared to brute speed. "Performance should be seen as being much more holistic than just speed," he said. "Simple development oversights can have massive implications."
Evidence of this is thrown up by the survey, which shows that over half the respondents encounter unexpected performance problems in at least 20 per cent of applications deployed. The most common problem this then causes is late delivery of applications – as reported by 66 per cent of respondents.
Late delivery, as Allen pointed out, only comes about when the performance problems show themselves immediately. However, many performance problems come in the form of slow degradation that manifests itself as a real issue some time after deployment. ®
Sponsored: Middleware for the modern age