|
|
|
|
Back to newsletter 275 contents
The Alibaba article I reference and extract tips from this month (see the articles section below and this month's linked tips page) suggests that much of JVM tuning is about garbage collection tuning. This reminded me of a finding one of my colleagues made when I was at Expedia which confirms that Alibaba approach.
My colleague analyzed production incidents at Expedia (for context, Expedia Group covers multiple brands including Expedia, Hotels.com, Vrbo, ebookers and many more, and has many thousands of engineers and many thousands of JVM instances running their systems). His analysis looked at incidents happening over 2 years across Expedia. Approximately 15% of incidents were resolved by changing the JVM configuration. Note that almost none of the incidents were actually caused by the JVM configuration, mostly these incidents were SLO failures. But by changing the JVM configuration - tuning the JVM - these failures were fixed. The particularly interesting part is that most of these - 13% the total incidents (ie 13/15=87% of these JVM configuration resolved incidents) - were fixed with GC changes. So the finding there was that the vast majority of JVM configuration changes needed in production are GC changes.
Now on to all the usual newsletter list of links, tips, tools, news and articles, and as usual I've extracted all the tips into this month's tips page
Java performance tuning related news
Java performance tuning related tools
Back to newsletter 275 contents
