JProfiler Get rid of your performance problems and memory leaks! Modern Garbage Collection Tuning Java Performance Training Courses COURSES AVAILABLE NOW. We can provide training courses to handle all your Java performance needs Java Performance Tuning, 2nd ed The classic and most comprehensive book on tuning Java Java Performance Tuning Newsletter Your source of Java performance news. Subscribe now! Enter email: Training online Threading Essentials course JProfiler Get rid of your performance problems and memory leaks!

Back to newsletter 272 contents

https://www.youtube.com/watch?v=lVmIEhYNAfg
Java Profiling and Performance (Page last updated May 2023, Added 2023-07-30, Author Frank Delporte, Publisher Foojay Podcast). Tips:

• JFR includes sampling CPU profiling; allocation profiling; safepoint profiling; and execution tracing.
• Profiling can also be used for diagnostics as well as performance engineering - finding unexpected issues rather than looking specifically for where to improve performance.
• For profiling, it's best to decide what you are looking for, what you are optimizing before you profile, so that you use the correct type of profiling.
• If you are tuning, you need to compare specific measures before and after, otherwise it's very misleading.
• The JVM contains ergonomics which tries to adapt and optimize the performance for the specific application being run.
• Distributed tracing is now massively important as it gives you the context of where and why a particular trace has issues.
• Performance tuning can be a cost optimization as well as a throughput or latency improvement.
• 83% of JVM apps don't even try to size the heap. Actually having no flag is better than the wrong setting of a flag.
• Algorithms and data structures still matter now, but most other inefficiencies are compensated for by the JVM.

https://medium.com/expedia-group-tech/essential-application-monitoring-metrics-a08519ecab9d
Understanding Real-Time Application Monitoring (Page last updated June 2023, Added 2023-07-30, Author Ritesh Kapoor, Publisher Expedia). Tips:

• Monitoring performance in real-time is critical for ensuring positive experiences for customers.
• Capturing the right metrics gives powerful insights into your application. Top performance indicators for requests are: TP99/TP95/TP90 percentiles, where TPx latency is the maximum time under which x% of requests have been served; average time taken; Maximum / Minimum time taken; Overall transactions processed per second.
• SLI/SLO/SLA should be defined and monitored for all applications. SLI=Service Level Indicator, a metric used for the SLO; SLO=Service Level Objective, target value or range of acceptable values for SLIs; SLA=Service Level Agreement, a (explicit or implicit) contract with your users that includes consequences of meeting or missing SLOs.
• Useful JVM metrics: Heap Memory Utilisation for different heap areas like Young, Eden space, etc; Major/Minor GC frequency; Major/Minor GC pause time; Total thread count.
• Useful application metrics: Response times, request count, request success rate, error status codes count, SLA breach count.
• Useful application datastore response metrics: response time per type of data request; timeout/connection reset count, queries per second, CPU and memory utilization of the datastore, page faults.
• Useful queue metrics to monitor: consumer consumption rate, producer production rate, average message size, producer-consumer lag.
• Useful cluster metrics: request count, response time, healthy host count, instances stopped per unit time.

https://www.youtube.com/watch?v=wApqCjHWF8Q
Secrets of Performance Tuning Java on Kubernetes (Page last updated October 2022, Added 2023-07-30, Author Bruno Borges, Publisher Devoxx). Tips:

• The most common fix for a performance issue is to increase the heap size or the number of replicas.
• The default choice of garbage collector algorithm (ie if the algorithm is not explicitly set on the command line) is the serial collector if there is 1 CPU available (including for containers that have less than 1000 millicores) is the serial collector; it is also the serial collector if less than 1.792GB is available; otherwise it is G1 after for Java 11+, or parallel for Java 8.
• JVM default heap size allocation is 50% of available RAM up to 256MB, then 120MB heap size until 512MB, after which it is 25%.
• Poorly tuned GCs can lead to high pause times, thread starvation and OOMEs.
• Selecting the appropriate garbage collector (GC) depends on the CPUs available - and similarly to attain specific GC targets you may need to provide more CPUs and a different GC.
• If running on bare metal, set the heap to whatever size the application needs, but less than available RAM; for a container set to less than 75% of container memory. Set using -Xmx or -XX:MaxRAMPercentage=75 - but note applications and 3rd party products using off-heap memory may need this lower.
• Recommended GC choices: serial for 1 CPU, parallel up to 2 CPUs and 4GB, G1 up to 4 CPUs, then Z or Shenandoah.

Jack Shirazi Back to newsletter 272 contents