I suspect that Cassandra has some needs similar to HAvroBase: (http://www.javarants.com/2010/06/30/havrobase-a-searchable-evolvable-entity-store-on-top-of-hbase-and-solr/).
Note, much of that project can apply to other key/value stores, its fairly abstracted from HBase in the key components like schema management. -Scott On Nov 27, 2010, at 10:59 AM, Eric Evans wrote: > > There is some work being done to compare the performance of thrift, > avro, msgpack, and cassandra's own hand-rolled message serialization. > > https://issues.apache.org/jira/browse/CASSANDRA-1765 > > Any input from the folks on this list would be appreciated. > > In particular, I'm curious about the erratic serialization times I'm > seeing from Avro, min vs max across a series of test runs, (see my > comment on CASSANDRA-1765 from 27/Nov/10 01:39 PM). > > -- > Eric Evans > [email protected] >
