A leading marketing software platform for the fitness industry needed a tune-up. The platform provides marketing tools and billing partner integration to any size business from stand-alone studios to enterprise organizations. Some dashboards on the web app were taking over 5 minutes to render. That’s not acceptable. Chariot’s goal was to dramatically reduce response time through back-end performance improvements.
Discovery & Determination of a Software Slowdown
The fitness marketing automation platform is cloud based. To identify slow performing calls, Chariot’s software engineers created a test suite on Amazon Elastic Compute Cloud, or EC2, to measure performance and run diagnostics on API calls.
The team established a baseline to measure the effectiveness of every change. Chariot used Apache JMeter, an open source tool, to identify issues within the overall back-end API implementation.
Over the course of a few weeks, engineers tried out potential solutions and measured effectiveness within the performance test suite. With concrete data in hand, the team was able to recommend specific changes to optimize back-end performance without increasing cost or adding complexity.
Optimized for Speed
Chariot made three major changes. First, the team added roll-up tables to hold summary data across many records in the database. Next, new database indices supported more finely grained dashboard queries. Finally, parallelizing operations meant that data collection requests could come from multiple sources simultaneously.
The result was a dramatic reduction in load levels for each of six dashboard configurations. For example, when 30 employees used the software simultaneously, load time in the scaled-down test environment went from an average of 22 seconds to just 2 seconds. At the more complex location level, when 50 users were on the platform at the same time, load time dropped from 129 seconds to 5 seconds on average. And at the club level, average load time was reduced from 346 seconds to 5 seconds.
Cloud Based Software Benefits From Tuning
While it may be tempting to address cloud performance problems by adding more nodes or increasing node power, there are real benefits to classic performance tuning.
In the case of the Club OS software platform, improvements to the application code, combined with improvements to the database, produced dramatic performance gains, with speeds up to 70 times faster. This kind of boost would have been extremely costly (if not impossible) to achieve by simply scaling nodes in the cloud.