Recently, there has been a lot of fun discussion about the death of SOA. Well SOA successes are quietly happening every day, and one of the keys is the use of robust testing and validation and, some cases, the use of service virtualization to speed up the testing and validation process.
SOA's shine may have worn off for some, but it is just another practice for designing and building dynamic, distributed enterprise apps. SOA may share boundaries with other technology practices: BPM, integration suites, and ESBs. All of these describe systems that are heterogeneous, distributed, multi-tier, and subject to frequent change by nature.
So call it what you will, not meeting the new challenges raised by this class of apps can lead to various unintended consequences and failures, or create bottlenecks to agility and cost savings which were the ultimate goal. We want to add to the conversation with a series of actual examples from the field to show successful cases of dynamic enterprise apps. This first one is a national cable provider.
This company had a very complex provisioning system - literally the "brains" that enables or disables what programming subscribers can see on their TV, depending upon requested configuration, on-demand status and special offers. The application was a complex application with several components that needed to interact at runtime in a secure transaction environment, and they had to operate efficiently and accurately to maintain customer satisfaction.
Our iTKO architect involved in this effort, Ken, said that before he and our integration partners at pureIntegration arrived, the cable provider staffers were hand building manual tests to try to cover these complex processes. Even a simple change could take the staff three weeks to build the less than adequate manual tests. Just getting ready to attempt to test a major release could take them several months, so the cable company was limited to only two major releases a year. This put them at a competitive disadvantage and added unnecessary expense to the testing and validation process.
Moreover, the QA manager noticed that manually testing was like trying to do deep sea exploration from a glass bottom boat. Gaining end-to-end test coverage and control, with automation, enabled them to move out of their testing silos and see the entire process. Now they can quickly spot error sources and correct them.
Ken said that in one case dozens of reported bugs were found to result from a single change. Getting past the 80%-20% rule and covering the most important and likely risk areas with automated testing first enabled them to spot these dependency failures in almost real time with each change. In the past, it was taking over three weeks to isolate the problem. Ken said that LISA was employed, enabling them to increase test automation and validation coverage from 15 – 20% to over 95%.
Ken added that “Saving several weeks in each test process has created a lot of excitement.” The results included the reduction of regression test execution from 3 weeks to 5-8 hours. The major release cycle was reduced by 6 weeks. The integration of LISA with the test and validation process also saved more than 95% of per-test costs, which is also an incentive to increase the test coverage to new features.
To read the entire case, go to iTKO LISA Case Study: Major Cable Service Provider.
Our next installment of this series will cover how a leading financial institution eliminated a cumbersome and expensive manual testing process, freeing up resources and dramatically decreasing implementation time.
Comments