We’ve talked a lot about how ensuring quality is a big challenge for complex SOA and enterprise software environments. But there is a bigger-picture risk that can drive any project past deadline and over budget: constrained and costly access to necessary systems and data throughout the development and testing lifecycle. This fifth example is about a major airline.
Like many companies, they faced constraints in their testing environment. In the airline industry, this is magnified, as all carriers have to make use of third party reservation systems. The billing for this access runs just like cell phone billing (but on a much bigger pricing scale!), with a set limit, followed by overage charges. Anything over the limit can get very expensive. In addition, the access itself is not cheap to begin with.
This airline wanted to expand their testing lab environments from four to nine. However, given the transaction fees I just described, and the associated hardware cots, this change would run into millions of dollars of extra costs on an ongoing annual basis. In addition to the extra transaction costs, the airline would have to purchase more hardware. Since you cannot just deal with airline and customer data over the public web for very real security purposes, they needed to provision special gateways and data circuits to interact with the third party source.
However, this was just the beginning of the problems they faced. Travel data is very complex and changing. There are weather delays, fare changes, rewards programs, partner systems, flight operation systems, and so on. To build applications, you need access to stable customer and flight data scenarios for testing and validation.
The airline found that when they ran regression tests, about 10% of the test cases failed. The failures were not mostly due to bad software, but data volatility. They had to find a way overcome a huge number of false negatives due to bad test data. Teams would often waste half their day chasing down these errors. Increasing the number of test labs by over 100% would only compound these data troubles.
Introducing a strategy that covered both automated testing and Service Virtualization addressed these issues, so we brought LISA to bear and demonstrated how to virtualize these kinds of services that are prohibitively costly and complex to replicate using conventional means.
First, they eliminated the need to incur almost all the transaction costs to the third party system by accurately simulating the required data as virtual services for development and test purposes.
This also took the additional hardware costs for gateways and data circuits out of the equation by reducing infrastructure needs and virtualizing any new infrastructure requirements. These hard dollar cost savings more than covered their investment in automated testing.
There was more, as the cycle of false negatives due to bad test data was broken, as stable test data could be provided. This may prove to be the largest source of saving overall. Now application development could proceed at a much faster clip, helping the airline generate customer benefits and revenue from the new application much quicker, and reducing development costs.
They are also planning to make use of the automated testing system and the data it generates to streamline their training programs and make them accessible on a local basis, reducing the travel expenses associated with getting their employees flying on the new systems they develop.
You can find the complete case study here.
Comments