> You mean the bandwidth between cluster? Yes. It should have big enough 
> bandwidth between them. The 2 clusters are in the same "cluster". I just 
> created 2 hbase/hadoop clusters.

Excellent.

> How do I pre-split? Can I do that using hbase shell?

That MR job is basically a massive read + write, so tunings for batch
loads apply: 
http://hbase.apache.org/book/perf.batch.loading.html#precreate.regions

It might just be that all the mappers were slamming a single region
server with all the writes, which is why I ask if when running it
again it's faster.

J-D

Reply via email to