Thanks for your reply. After I tried it on my PC it became obvious that it is a disk issue.
Gary Tully wrote: > > That sounds like a disk limit as each send/commit will cause a file sync > and > the client is blocked till that completes. With batching, things improve > as > the disk writes coalesce. > > It may be interesting to run this little disk benchmark that replicates > the > kahaDB writer thread to quantity the effect of the sync on some difference > machines. > > java -classpath lib/kahadb-5.3.1jar org.apache.kahadb.util.DiskBenchmark > > > On 27 April 2010 15:30, jid1jid1 <j...@dharma.gr> wrote: > >> >> I use ActiveMQ 5.3.1 with default configuration. >> My message producer rate for 1 msg / per commit cannot exceed >> 90msgs/second. >> >> I tried using multiple Thread/Sessions, Connections and JVM instances and >> the rate cannot exceed this (without settings like >> factory.setUseAsyncSend(true) ) >> >> The weirdest part is that I get the same 90msgs/sec even in every >> possible >> combination and even if I try sending messages to different queues. >> >> If I change the usual setting like committing sessions in batches or >> factory.setUseAsyncSend(true) I go up to 5000 msgs/sec. >> >> (BTW, I get the same slow performance even independently if I run my >> producing application from the same physical box or remotely). >> >> Any ideas of what to look for? >> >> Many thanks >> -- >> View this message in context: >> http://old.nabble.com/Weird-producer-slow-performance-%28with-persistence%29-tp28377685p28377685.html >> Sent from the ActiveMQ - User mailing list archive at Nabble.com. >> >> > > > -- > http://blog.garytully.com > > Open Source Integration > http://fusesource.com > > -- View this message in context: http://old.nabble.com/Weird-producer-slow-performance-%28with-persistence%29-tp28377685p28385845.html Sent from the ActiveMQ - User mailing list archive at Nabble.com.
