svn commit: r1592814 - in /kafka/site: 081/ops.html images/kafka_logo.png

[ne...@apache.org]

Author: nehanarkhede
Date: Tue May  6 17:49:08 2014
New Revision: 1592814

URL: http://svn.apache.org/r1592814
Log:
Java 7 GC settings

Modified:
    kafka/site/081/ops.html
    kafka/site/images/kafka_logo.png

Modified: kafka/site/081/ops.html
URL: http://svn.apache.org/viewvc/kafka/site/081/ops.html?rev=1592814&r1=1592813&r2=1592814&view=diff
==============================================================================
--- kafka/site/081/ops.html (original)
+++ kafka/site/081/ops.html Tue May  6 17:49:08 2014
@@ -354,17 +354,21 @@ producer.purgatory.purge.interval.reques
 Our client configuration varies a fair amount between different use cases.
 
 <h3><a id="java">Java Version</a></h3>
-Any version of Java 1.6 or later should work fine, we are using 1.6.0_21.
+We're currently running JDK 1.7 u51, and we've switched over to the G1 collector. If you do this (and we highly recommend it), make sure you're on u51. We tried out u21 in testing, but we had a number of problems with the GC implementation in that version.
 
-Here are our command line options:
+Our tuning looks like this:
 <pre>
-java -server -Xms3072m -Xmx3072m -XX:NewSize=256m -XX:MaxNewSize=256m -XX:+UseParNewGC -XX:+UseConcMarkSweepGC 
-     -XX:+UseCMSInitiatingOccupancyOnly -XX:+CMSConcurrentMTEnabled -XX:+CMSScavengeBeforeRemark 
-     -XX:CMSInitiatingOccupancyFraction=30 
-     -XX:+PrintGCDetails -XX:+PrintGCDateStamps -XX:+PrintTenuringDistribution 
-     -Xloggc:logs/gc.log -Djava.awt.headless=true
-     -Dcom.sun.management.jmxremote -classpath &lt;long list of jars&gt; the.actual.Class
-	</pre>
+-Xms4g -Xmx4g -XX:PermSize=48m -XX:MaxPermSize=48m -XX:+UseG1GC
+-XX:MaxGCPauseMillis=20 -XX:InitiatingHeapOccupancyPercent=35
+</pre>
+
+For reference, here are the stats on one of LinkedIn's busiest clusters (at peak):
+        - 15 brokers
+        - 15.5k partitions (replication factor 2)
+        - 400k messages/sec in
+        - 70 MB/sec inbound, 400 MB/sec+ outbound
+
+The tuning looks fairly aggressive, but all of the brokers in that cluster have a 90% GC pause time of about 21ms, and they're doing less than 1 young GC per second.
 	
 <h3><a id="hwandos">6.4 Hardware and OS</a></h3>
 We are using dual quad-core Intel Xeon machines with 24GB of memory.

Modified: kafka/site/images/kafka_logo.png
URL: http://svn.apache.org/viewvc/kafka/site/images/kafka_logo.png?rev=1592814&r1=1592813&r2=1592814&view=diff
==============================================================================
Binary files - no diff available.