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 148 contents

A colleague described to me what happened when he was setting up a booth at a small IT show. During preparation, as the booths slowly turned on their equipment, at some point the circuit breakers tripped and power was lost to all equipment. It turned out that there was a misconfiguration of the power equipment, and after twenty minutes, the problem was identified and the circuit breaker re- enabled. But the circuit breakers immediately tripped again. After trying a few more times and having the same thing happen, the power engineer searched for the new problem in the power equipment. Everywhere he probed, there was no detectable problem. But whenever he re-enabled the circuit breaker, it immediately tripped it again. Finally, he worked out that because all the booths were almost fully plugged in, as soon as the circuit breaker was re-enabled, the equipment in all the booths simultaneously tried to fire up causing a huge surge that the power equipment couldn't handle, and so causing the circuit breaker to trip (correctly protecting the power providing equipment). He actually had to disconnect the majority of booths manually, and then gradually reattach them to the circuit before the floor was live again.

A note from this newsletter's sponsor

I'm sure you can see the analogy here. Some of you have probably even seen it happen to your servers. Something tips your server over the edge, causing it to fail. You bring it up as quickly as possible - maybe it is even redundant and you have an automatic failover to your reserve system. Then the restarted or reserve system immediately fails too. The cause is from all the subsidiary systems or users reconnecting. If it's an outward facing webserver, users are refreshing their pages because the page didn't display so you get an immediate storm of requests. It's a particularly embarrassing failure mode, because the more successful the system, the more likely this is to happen. Load tests and stress tests rarely check for request storms on startup or failover, because the standard testing paradigm is to ramp up load gradually since that tends to be the most common load increase. If you don't have the "request storm on startup and failover" scenario in your load test performance suite - it's time to add it.

Now on to all our usual links to Java performance tools, news, articles and, as ever, all the extracted tips from all of this month's referenced articles.

A note from this newsletter's sponsor

News

Java performance tuning related news.

• A reminder about CMS phases from one of the current Oracle GC guys
• CMS logging output explained by one of the current Oracle GC guys
• GC analysis using Censum - note the valid point that GC logs are still the most useful way to analyse GC, as the JMX information is limited and can itself create extra garbage
• "You don't build a bridge, then try to add load-bearing capabilities at the end of the project -- but most software projects try to do exactly that." Lovely quote describing why most projects don't really do the "engineering" part of "software engineering"

Tools

Java performance tuning related tools.

• Striped Executor Service, allows you to "stripe" execution in such a way that all tasks belonging to one stripe are executed in-order
• Perfidix is a tool for developers to benchmark Java code in JUnit style
• LightFish opensource application which periodically gathers and persists GlassFish's monitoring data
• Java thread local profiler

A note from this newsletter's sponsor

Articles

• Taming the Performance Beast - a Practitioner?s Way (Part 1) (Gopal Sharma/DZone)
• How to Analyze Java Thread Dumps (Tae Jin/DZone)
• High-speed, multi-threaded virtual memory in Java (Alexander Turner/jaxenter)
• Integration performance tuning for WebSphere Commerce and Sterling Order Management (Charek Chen, Vijaya Bashyam, Lei Sun/IBM)
• Application performance manager: Making the most of limited resources (Jennifer Lent/SearchSoftwareQuality)
• Scalability lessons from Google, YouTube, Twitter, Amazon, eBay, Facebook and Instagram (Dodgy Coder/Dodgy Coder)

Jack Shirazi

Back to newsletter 148 contents