Works Applications Co., Ltd.

"For example, suppose a customer informs us that performance degraded after a new feature was delivered. To identify the cause, we had to start by interviewing customers about the usage status of HUE. We used various methods to collect information, such as Oracle Statspack, Java thread dumps, and DB queries. We thoroughly inspected the contents, then searched for the causes of the delay,” explains Hayato Shiraiwa, AC SRE Department Manager for Works Applications.

Even if the team modified the source code to improve performance, it might take a week or two for the issues to be resolved. When various investigations did not resolve the problems, they sometimes had to add more infrastructure resources at an extra cost.

Furthermore, many performance degradation problems had been discovered through customer feedback. There was a growing need internally for the development and SRE teams to be able to detect problems by themselves.

"For us to transform into a cloud service vendor, we have believed from the beginning of HUE's development that we should target a mechanism that could proactively address problems as they arise," says Masahiko.

Before the full-scale implementation of AppDynamics, Works Applications conducted Proof of Value (PoV) to confirm actual usability. "When I first used AppDynamics, I was surprised that it could display function-level performance without any configuration. I could check the performance of the customer's system usage without having to go out and get data," Hayato recalls. The AppDynamics PoV was highly regarded for its ability to visualize detailed information, such as how many milliseconds each Java process took and why a particular code stopped processing.

AppDynamics greatly reduced the MTTR across the platform

The implementation of AppDynamics has made it easier to identify the causes of problems, including application performance degradation, resulting in various positive results. The SRE department’s time spent resolving issues was reduced to about one-third to one-half of what it used to be. This allowed the SRE department to focus more on automating its infrastructure's operation and maintenance and improving its service quality. 

"AppDynamics centrally and automatically accumulates customer DB and application performance metrics and traces, eliminating the need to collect various levels of logs manually. Simply drilling down in the AppDynamics dashboard allows us to identify the causes of delays easily. In some cases, the task of identifying causes, which used to take nearly a week to identify, can be completed within minutes,” says Hayato. "Additionally, the SRE department is now able to quickly respond to questions and inquiries based on the root cause analysis with evidence such as correlation of application delays to a problem with infrastructure performance and provide clear answers regarding the causes.

When a certain threshold is exceeded, AppDynamics can automatically send notifications about monitored metrics. The company sends these notifications to stakeholders via chats. This helps developers and the quality control department to detect problems immediately and address them proactively.

In this way, AppDynamics has been helpful for operation management, such as monitoring applications and DBs and identifying the causes of performance degradation. "This is not just a monitoring tool. We want to analyze the data stored in AppDynamics and use it to improve our business, such as by determining how to grow our services and how to use it to solve customers' business problems," says Hayato. Furthermore, Masahiko adds, "The impact of the current implementation is merely a passing point. We aim to move from simple monitoring to business observability to provide new value to customers."