Generally, Rapleaf has found it to be an extremely stable and reliable server. The only gotcha I'm aware of is https://issues.apache.org/jira/browse/THRIFT-714, which just means that your total number of worker threads will not vary quite as much as you think it might based on load. (I'd love someone to contribute a real fix for this problem...)
Another thing to be aware of in general is https://issues.apache.org/jira/browse/THRIFT-786 - you might not get the I/O performance you're expecting if the actual RPC I/O is large. This also could be fixed with relative ease if it turns out to be a problem. -Bryan On Wed, Jul 7, 2010 at 9:04 AM, Jonathan Ellis <[email protected]> wrote: > Cassandra is looking at switching from TThreadPoolServer to > THsHaServer to reduce the amount of memory required by large numbers > of client connections. Are there any gotchas we should be aware of? > Early testing suggests performance is competitive. > > -- > Jonathan Ellis > Project Chair, Apache Cassandra > co-founder of Riptano, the source for professional Cassandra support > http://riptano.com >
