Say you’ve got a big application upgrade project, or a big server consolidation project. Projects like these require a sound best practice for virtualization capacity planning. CA’s acquisition of Hyperformix centers on improving CA’s virtualization management – helping IT perform capacity planning and right-sizing production resources while ensuring acceptable application performance. At Shunra we call this “deploying with confidence”.
Bruce Milne of CA provides a great overview of how Hyperformix helps IT meet this challenge:
http://www.hyperformix.com/resources/videos/hyperformix-101
A key takeaway is that CIOs are leaving $millions on the table by utilizing < 20% of their server resources in production. A key reason is they lack confidence in their ability to accurately predict production resource requirements given normal business growth, peak period load, seasonal cycles, etc. So, they pay for headroom – way too much.
In the Hyperformix Capacity Planning best practice, the “Test to Production” phase takes data from your test lab, applies these results to a model of your production network architecture, and then predicts your performance in production. From Hyperformix Web Site:
“Turn Load Testing into Insight – Load testing is a critical component of mission critical product releases. HP Performance Center and LoadRunner users can leverage the data generated by load testing into models that accurately portray how the application will behave in production — no need to be concerned that the test lab is different than production. Optimizer will identify bottlenecks and illustrate parts of the infrastructure that are overbuilt and can be consolidated.”
A critical flaw of the “Test to Production” phase can occur if you use load test results that are not based on tests that include your production network behavior. The response times and server resource consumption patterns of 1000 “virtual users” in a lab vary significantly compared to 1000 users accessing apps from various network locations while competing with bandwidth limits, latency and occasional packet loss. We’ve seen system scalability drop from 1000s of users to 100s of users, and response times rise from several seconds to several minutes, when apps move from a test lab into production. Given the limited accuracy of the current best practice it’s no wonder CIOs pay for lots of headroom.
The good news – HP has solved this problem with the Shunra for HP Software product. The “Test to Production” phase can now include load testing that easily incorporates the production network behavior of your remote locations – and the solution is built-in to HP’s Controller and Load Generator. Shunra is offering a free 30-day trial license that turns on this feature inside LoadRunner (Version 9.5 & above). You’ll be surprised when you see the significant difference in your test results accuracy. You’ll also gain more confidence in reclaiming some of the 80% of unused server capacity sitting in your data center.
Leave a Reply