Hi, I know no-one can give a precise figure. But I wondered what reasonable write throughput (on a single Region Server writing into a single region, ignoring compaction) folk would expect relative to HDFS, given you would anticipate HBase becoming WAL-bound at it's max speed?
E.g. if you can achieve 200MB/s max throughput on HDFS what would you broadly expect to achieve on HBase on the same node, given the write-path includes both WAL and memstore updates? Cheers, Owen -- Owen Rees-Hayward twitter.com/owen4d
