EAI projects suffer from a failure rate of roughly 70 percent, and attendees of an EAI Industry Consortium workshop in July had little trouble sharing and detailing numerous “bear traps” that projects had been tripped up by, causing them to run late, go over budget, produce less-than-hoped-for results, or any combination of the three.
“They're the problems people generally fell into and were severely hurt by, so we thought 'bear traps' was a reasonable name for them,” EAIIC European Chairman Steve Craggs explained.
During the ebizQ webinar Avoiding the Pitfalls of an Integration Project , part of the Best Practices for IBM WebSphere Software series, sponsored by Candle Corp., Craggs and Candle Senior Architect Peter Rhys Jenkins told how to achieve the exacting levels of coordination needed between and beyond business and IT units to sidestep these pitfalls.
“None of them are really technical. Almost all are management issues,” observed Craggs, who before founding Saint Consulting Limited was the worldwide executive in charge of IBM’s MQ Series. “Most of these might sound fairly obvious, but it's amazing how many companies have fallen into these problems, either by not taking account of them at all or perhaps taking an inadequate account of them.”
The seven main bear traps brought up by consortium members, and their possible solutions as described by Craggs during the webinar included:
1. Change is constant: “EAI implementations are very fluid; they change frequently, require changes in different components and what's more, they spread across businesses and even across value chains into different companies as a business process integrated all the way through from start to end.”
The common practice of budgeting to the end of a particular project gets companies in trouble when service level and load requirements increase after deployment -- so companies should create a post-project investment stream that allows for more than just basic maintenance.
2. EAI skills are rare: EAI’s complex (and often proprietary), asynchronous parallel processing and data combinations can befuddle programmers and help-desk staffers who are used to a straight-line approach.