On 5/11/2013 5:56 AM, Spadez wrote: > Tomcat Vs Jetty > Open JDK vs Oracle JDK > > I haven't been able to find any information online to backup any sort of > performance claims. I am planning on using Tomcat with Open JDK, has anyone > had any experience with this and is it a wise path to go down?
I use the stripped-down Jetty that's included in Solr. I have upgraded it to a newer 8.1 point release than what's included in the 4.x download, but that's it. For my remaining 3.5 servers, I am using the patched Jetty 6 that's included in the download. I use Jetty over another container for a few main reasons: - It's very lightweight. -- Lighter than a full jetty install. -- Definitely lighter than tomcat. - The config has been lightly tuned for good results with Solr. - I don't have much experience with other containers. - It's used EXTENSIVELY in Solr's JUnit test framework. There have been a lot of bugs over the years along the lines of "This feature doesn't work in X, but it works in Jetty" ... and "X" is usually Tomcat. General problems and problems specific to Jetty are caught VERY quickly by Solr's test suite, but other containers don't receive much testing before release. As for JVM, I use Oracle. They (via Sun) created it, and they maintain it. It's the version that's least likely to have problems. As I watch the development list and Jira activity, I see more reports about bugs in the IBM and OpenJDK versions than I do about bugs in the Oracle version. IBM is far worse than OpenJDK. Lucene/Solr code is particularly good at tripping JVM bugs, and when they are found in the Oracle JVM, they usually affect all of the JVMs, not just Oracle. I'm sure there are exceptions, but I don't recall any at the moment. This next claim is entirely subjective, I have no concrete information either way: Oracle seems to put out fixes for major bugs faster than any of the others, including OpenJDK. I can't believe I'm saying that about something that's not open source. :) Thanks, Shawn
