Ramsey said Q2 initially tried monitoring applications by creating health rules based on a certain percent of errors, but that approach ended up being too noisy. To reduce false alerts, Ramsey switched to dynamic baselines—these are automatically created by the AppDynamics controller using machine-learning algorithms to account for changes to baseline performance over time. “By incorporating dynamic baselines we were able to drive down the number of false alerts by more than 83%,” Ramsey reported.
Another way that Ramsey ensures end users will have the best possible experience is to keep tabs on how the platform is performing for different customers. To this end, he autogenerates individual dashboards so each customer has the option of tracking their own performance.
Use of AppDynamics has reduced troubleshooting by hours—and in some cases days— by pinpointing the location of a problem. “It really helps to have a giant, blinking red dot to tell people where to start looking when you have applications spread over 100 different servers and a variety of different applications,” Ramsey said. “Rather than assemble folks from the networking team, storage team, VMware team, and app team in an IT war room where they would typically end up pointing fingers at each other, AppDynamics helps everyone get on the same page.”
Q2 has also seen daily error rates drop using AppDynamics and experienced an increase in overall code quality. “AppDynamics alerted us to a lot of existing problems that we didn’t know we had,” Ramsey said.
Like many organizations, Q2 is transitioning to DevOps and has already begun to increase release velocity. As teams become more agile, AppDynamics is giving them confidence that new releases will not introduce regressions. The assurance is not only important to internal teams, but also helps build trust with banks and credit unions who need to have confidence that each new version of the platform is highly performant before they turn it over to their customers. With AppDynamics, financial organizations can run their own load tests and observe the results themselves.