Looking for ways to improve transaction speeds and increase concurrent users? How do you troubleshoot an OutOfMemoryError and reduce memory footprint? (Answer: Hopelessly holed up in abilities lab trying various 'fixes'?)

Although timers and heap dumps indicate the existence of a problem, it is difficult to narrow down to the exact code location, especially for recursive or multi-threaded programs. Java Flight Recorder (JFR) is a free profiling tool which helps us take snapshots of JVM while our services run to analyze various performance metrics.

As MSVC Orders & Tasks team, we have utilized JFR to profile heavily used Millennium services owned by us such as Google-like medical search engine, e-Prescribing, and order retrieval. Join us as we walk you through several success stories from incorporating JFR in our development process. Examples include:

1. Enhance performance of multiple EJS services, with the best-case improvement of 9x in both memory footprint and transaction time.
2. Fix a concurrency issue, allowing one of our heavily used service to scale from (previously limited to) 5 to (now comfortably) 15 concurrent users.
3. Detect a memory leak that caused a new EJS to die after a few cycles.
4. Resolve a memory issue in test framework that caused VM crash on our Jenkins pipeline.

Learning Outcome:

1. Know how to setup Java Flight Recorder using JMeter or Gatling for various scenarios.
2. Drill down stack trace memory analysis on EJSs, tests, or any Java code in general during and post development for large memory footprints, memory leaks, and resource-intensive methods.
3. Pinpoint and resolve concurrency issues.

Target Audience:

Java developers and Java Service Owners looking to improve their services during and post development.