Hi, We've been setting up our architecture at work using HBase quite extensively but since we are developing in Python we have chosen to sue Thrift to converse with the HBase cluster.
Now it does look like you can only start one single Thrift server per machine? Could it be a bottleneck in comparison to writing directly into HBase when using many clients writing rather rapidly? Could I leverage that potential bottleneck by starting more than one Thrift server on a given machine (that would require I assume being able to pass a specific configuration to HBaseConfiguration in ThriftServer, right? Is there any potential feedback on the matter? I'm a little worried that Thrift cannot actually cope above a certain load but I lack experiences on the subject. Thanks, - Sylvain -- Sylvain Hellegouarch http://www.defuze.org
