avro-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Leo Romanoff <romix...@yahoo.com>
Subject Speed improvements for reflection-based serialization
Date Thu, 21 Mar 2013 07:39:30 GMT

I've played a bit with Avro serialization based on reflection, i.e. with the classes from org.apache.avro.generic
It works fine in this mode, but is rather slow compared to such frameworks like protostuff
or kryo.
Quick look at the source code has shown that a lot of reflection-based operations and class
lookups are not cached. And such operations are usually pretty expensive when executed by
a JVM. 

So I changed some of org.apache.avro.generic classes, introduced caching and a few other optimizations.
Now it seems to perform much better.
I could submit my patch for a review, if anyone is interested in such improvements. I'm new
to Avro, but I got the impression that Voldermont and may be a few other BigData projects
are using it? May be they are interested, though I don't know if they use this reflection-based

Best Regards,

View raw message