Question
1
Replies
131
Views
Posted: July 31, 2017
Last activity: August 11, 2017
Closed
DNode RF settings for clusters of 6 nodes
All,
We currently use the default RF of 3 for Cassandra for the data keyspace in a 6 node cluster. Has any played with changing this value in larger clusters?
***Updated by moderator: Lochan to add SR Exists group tag***
I didn't and I wouldn't.
If you increase the replication factor, then more nodes need to be contacted when you do a "write" for instance. So it is going to be chattier in the network. There would be more "write" overall in the cluster so the cluster will be slower. You also lose in terms of space. You only win in terms of availability.
For instance we a consistency level set to 1 and a RF of 5, the data will still be available in 1 node even if 4 are failing.
If you decrease the replication factor then the opposite happens.
RF of 3 appears to be the right middle ground.