Hey all, I have a small cluster of HBase, Hadoop HDFS, Zookeeper that is being deployed on the same nodes (kind of a common best practice for small clusters). Everything is quite straight forward and works fine.
I am wondering how can I efficiently predict the right time to add another node from the HBase performance (read/write) perspective ? Is there any suggested amount of regions per region server I should follow? number of requests per region service? any other "load" parameters? So as long as the data keeps increasing, I guess performance will be decreasing. moreover. I find it hard to answer the question of "when should I add another Hbase node?" because performance decrease can be caused by other reasons like "hots pots" for example, which demand a special care what not another node. Suggestions? Thank you all!! -- * Ronen Itkin* Taykey | www.taykey.com
