Semiconductors fuel the growth of mobile devices, wireless networks and defense technologies. Supplying these innovative chips requires rapid design and manufacturing processes, as well as predictable and scalable application performance.
When Qorvo’s applications slow down or stop, so does its production line. As Aparajit Saigal, manufacturing IT team lead at Qorvo explains, “Performance hindrances to any one of our applications can cause a bottleneck in the fabrication plant or on the shop floor, resulting in late delivery, lots of overtime, and lost revenue.”Qorvo’s semiconductor wafer fabrication facilities depend on a myriad of .NET applications that tie engineering systems to the manufacturing execution environment that drives wafer fabrication on Qorvo’s shop floors. A daily stream of code updates flows through a continuous integration cycle that feeds briefly into alpha servers. Once code passes performance and stability tests, it’s placed directly into production.
Qorvo’s core application, called “Nugget,” comprises multiple web services, several types of databases (Oracle, MySQL and SAP), and a user interface. In addition, each database is populated by other applications from legacy systems written in multiple languages on different architectures, all connected by a darknet overlay network.Prior to deploying AppDynamics Application Intelligence, Qorvo did not have an application performance management platform. Saigal explains:
to pinpoint performance bottlenecks. We were searching through logs and doing a lot of analysis. Then, we started trying to write our own monitors at different levels of the application. The problem with this approach is that you’re spending time writing bloated code, which will require additional analysis to hopefully determine a performance issue.”
As Qorvo continued to grow, application performance became a bigger issue. “The application simply couldn’t scale,” recalls Saigal. “What’s more, we didn’t have the in-house expertise to figure out what needed to happen in order to scale the platform.”
With any application, users first notice performance issues at the user interface—regardless of the actual root cause. More often than not, performance issues involve database queries.
During a recent code release, Saigal’s team noticed performance in a core database that drives production applications had slipped from a second and a half to six or seven seconds. However, resolving the issue was easy. AppDynamics Application Intelligence showed exactly when the slowdown happened and which code was involved. This allowed Saigal and his team to quickly revise the code and fix the problem without impacting other database processes.
Qorvo has found the AppDynamics Application Intelligence Platform to be especially useful in pinpointing delays in Active Directory. In another instance of performance degradation, in which DevOps made a small change occurring between two Active Directory servers, page views slowed from around two seconds to 15 or more.
“As soon as we opened up AppDynamics,” says Saigal, “the cause of the bottleneck became obvious.” He adds, “I can safely say we’ve reduced interruptions to production by at least 60 to 70 percent thanks to the AppDynamics Application Intelligence Platform.”
But perhaps Saigal’s concluding remarks are the best testament of all: