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 181 contents

99.99%. That's the new "standard" target to achieve. Well, not really, it all depends on your needs and your customers' requirements, but there's some buzz around that "Four 9s" target.

A note from this newsletter's sponsor

Another way to think about it is that a 99.99% success rate is one failure allowed in 10,000. Over a day (86400 seconds) that would mean you achieve your targets in every second of the day except for about 8 or 9 seconds. If you were talking about uptime, that means you have less than an hour's total downtime over the whole year (8760 hours). For reference, Facebook, Apple, Twitter, Netflix, and many more huge sites that need to provide 24/7 sevice failed that "less than 1 hour of downtime" target in 2015; as did parts of Google (eg DoubleClick).

Most targets aren't simple to define. For example, if your page needs to download and display in under a second, does that apply every time? Any SLA which requires that is guaranteed to fail - networks are unreliable and you simply cannot guarantee that a remote request succeeds every time. But you can define a target such as it displays in under a second 99.99% of the time. Or more nines if appropriate to your application. At the moment, four 9s is the fad. Next year or the year after it will be five 9s - you might want to consider just what you need to change to achieve that. And start planning now.

Now on to our usual links to tools, articles, news, talks 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.

• Comparison of various garbage collectors for a particular application (G1 with -XX:+UseStringDeduplication wins)
• Latest changes for JEP 261, the Module System (still scheduled for Java 9, final release which has been pushed back to March 2017 because of that)
• Java 8 JVM escape analysis in action
• Ratpack high performance Java web framework first full version released

Tools

Java performance tuning related tools.

• ND4J scientific computing libraries, designed to run fast (supports GPUs) with minimum RAM requirements
• high performance Java web framework
• Java port of a concurrent trie hash map implementation from the Scala collections library (from https://github.com/axel22/Ctries)
• Using chrome timeline

A note from this newsletter's sponsor

Articles

• High Performance Lists in Java (Thomas Mauch/JavaZone)
• High-Concurrency HTTP Clients on the JVM (Fabio Tudone/Parallel Universe)
• Weak, Soft and Phantom references: Impact on GC (Gleb Smirnov/plumbr)
• Everything You Know About Latency Is Wrong (Tyler Treat/Brave New Geek)
• 5 Tips for Reducing Your Java Garbage Collection Overhead (Niv Steingarten/takipi)
• Which thread executes CompletableFuture's tasks and callbacks? (Tomasz Nurkiewicz/nurkiewicz)

Jack Shirazi

Back to newsletter 181 contents