Reply
New Contributor
Posts: 4
Registered: ‎02-20-2015

Best practices re: putting services with spiky CPU and memory usage on different servers from Hbase?

Are there any best practices concerning what services should be cotenants on the same cluster nodes? From what I understand about Spark, it's basically designed to fully utilize the available memory and CPU on the host machine, within certain limits. Our usage of Spark has been very spiky so far--periods of inactivity followed by the entire cluster being pegged for short periods. In contrast, Hbase is designed to perform as a database, with fairly consistent performance, lower patency, etc. Would you recommend putting Hbase and Spark services on different nodes so the CPU spikes from Spark don't impact the Hbase service? 

Highlighted
Posts: 1,896
Kudos: 433
Solutions: 303
Registered: ‎07-31-2013

Re: Best practices re: putting services with spiky CPU and memory usage on different servers from Hb

If the HBase latencies are SLA critical to you, we've often recommended separating out the HBase cluster to its own - this gives much more reliable performance and we indeed have multiple customers running it this way.

Otherwise, configuring your memory and CPU resource tunables on the cluster to divide it in a good way between the services is important.
Announcements

Our community is getting a little larger. And a lot better.


Learn More about the Cloudera and Hortonworks community merger planned for late July and early August.