I noticed that it says "Open source, supports file-based persistence.", while it is not true.
It would have been great if they provided the source used for the benchmark.
Also no mention about the version of jbossmq used.
?! what is not true about the above statement?
jbossmq does support jdbc persistence too.
The reason that the highlighted file based persistence is that they think the file based persistence implementations are faster since there is no Relational Database overhead.
Is this true?? Well the day I get myself off my lazy ass and do some performance testing of the file vs. jdbc persistence, I'll let you know.
this should be highly dependent upon the underlying RDBMS storage system. this holds true for the file-based PM, too, as the various implementations of filesystems (i.e. linux ext2, ext3, win fat32, ntfs, etc.) are very different from each other - i can imagine that one wins over the other when using rather large messages while it loses big time when you send a lot of very small messages.
per se there can't be a "Best" Persistence Manager which performs well for each and every application.
They took it off the web, because the benchmarks were biased and full of mistakes. So don't bother.
The best way to benchmark a JMS and to find the fastest for your purpose, is to build YOUR application according to the JMS specification and then "plug-in" different JMS servers.