Customer
How Bracket Computing Increased Throughput x10 by Load Testing and Continuous Integration
" BlazeMeter provides gorgeous reports that visualize the performance test in a way JMeter can’t do. By converting tests and running them through BlazeMeter with Taurus and Jenkins, we are able to constantly ensure our system is running as we want it to."
The Challenge
The Bracket Computing solution enables large enterprises to run their most sensitive workloads securely on the cloud. Bracket customers might run 10,000 instances against the system at a single time, requesting decryption keys.
The Solution
To check Bracket software’s handling of customer loads, Bob Van Zant, a software engineer at Bracket, uses the BlazeMeter solution for performance testing. With BlazeMeter, he simulates 10s of thousands of customer instances (instead of launching actual virtual machines) to test the system’s stability and robustness. He runs the tests every night through Taurus, an open-source testing automation system, and Jenkins, an open-source continuous integration system.
Afterwards, Bob analyzes the results in rich and visualized reports. He looks at different KPI correlations, follows up on system health trends and checks pass-fail criteria. He doesn’t do this alone - results are shared with his colleagues who can see the test results on the BlazeMeter site.
This process of continuous testing helps Bracket discover in advance if there are any problems that need fixing, to reinforce their confidence that the system will support all of their customers’ needs, all the time.
The Results
Thanks to the consistent and repeated load testing they conducted, Bob and his team were able to discover and fix two bottlenecks. First, they identified and fixed the way they were using their database. They discovered the problem when they noticed that the performance test generated low levels of concurrency when running on the backend.
Second, they discovered through Blazemeter load testing and application profiling that they were using massive amounts of CPU terminating TLS within the containers. Bracket’s approach includes using TLS for data-in-motion encryption. This information helped them improve solution scalability. Fixing both bottlenecks also served them financially by allowing them to cut the number of CPU cores in their baseline deployment by more than half, significantly lowering their spend with Amazon Web Services.
Bob found BlazeMeter after searching for a load testing solution. He started out with JMeter, but discovered that the enhancements BlazeMeter provides for JMeter, especially the reporting features and the collaboration opportunities, best suit his needs.
One of the reasons Bob chose BlazeMeter was the business model, which lets you sign up and immediately run tests. In addition, BlazeMeter supports SAML, which fits Bracket’s needs, and integrates with Okta, Bracket’s identity provider.