Re: mixed nodes, some SSD some HD

2014-03-06 Thread Robert Coli
On Wed, Mar 5, 2014 at 7:28 PM, Elliot Finley wrote: > What are the penalties for having them? > Vnodes are basically nodes, so the same penalties you would get from a huge number of nodes... folks have worked on optimizing the worst aspects of this, but the reality is still that a node with 256

Re: mixed nodes, some SSD some HD

2014-03-05 Thread Elliot Finley
What are the penalties for having them? On Wed, Mar 5, 2014 at 5:59 PM, Robert Coli wrote: > On Wed, Mar 5, 2014 at 2:53 PM, Tyler Hobbs wrote: > >> >> Ah, I missed that part. If the replication factor matches the number of >> nodes in the cluster, then you're correct, it doesn't matter what

Re: mixed nodes, some SSD some HD

2014-03-05 Thread Robert Coli
On Wed, Mar 5, 2014 at 2:53 PM, Tyler Hobbs wrote: > > Ah, I missed that part. If the replication factor matches the number of > nodes in the cluster, then you're correct, it doesn't matter what tokens > you assign. > And you get no benefit from 3x256 vnodes, but you do pay all the penalties fo

Re: mixed nodes, some SSD some HD

2014-03-05 Thread Tyler Hobbs
On Wed, Mar 5, 2014 at 4:48 PM, Elliot Finley wrote: > How does giving the SSD node a higher number of tokens help anything if > it's a 3 node cluster with 3 replicas? Ah, I missed that part. If the replication factor matches the number of nodes in the cluster, then you're correct, it doesn't m

Re: mixed nodes, some SSD some HD

2014-03-05 Thread Elliot Finley
How does giving the SSD node a higher number of tokens help anything if it's a 3 node cluster with 3 replicas? On Wed, Mar 5, 2014 at 3:41 PM, Tyler Hobbs wrote: > The dynamic snitch wraps whatever snitch you configure (SimpleSnitch, > PropertyFileSnitch, etc). > > The dynamic snitch *will* fa

Re: mixed nodes, some SSD some HD

2014-03-05 Thread Tyler Hobbs
The dynamic snitch wraps whatever snitch you configure (SimpleSnitch, PropertyFileSnitch, etc). The dynamic snitch *will* favor faster replicas, but it might still under-utilize the SSD nodes. Chris's suggestion to give the SSD nodes a higher number of tokens will result in them being replicas fo

Re: mixed nodes, some SSD some HD

2014-03-05 Thread Russell Bradberry
Are you using the dynamic snitch? Because the SimpleSnitch is the default. On March 5, 2014 at 5:27:03 PM, Elliot Finley (efinley.li...@gmail.com) wrote: Keep in mind, for this 3 node cluster, N = 3. I did a bit more digging and I found this (for future searches on this topic): http://www.dat

Re: mixed nodes, some SSD some HD

2014-03-05 Thread Elliot Finley
Keep in mind, for this 3 node cluster, N = 3. I did a bit more digging and I found this (for future searches on this topic): http://www.datastax.com/documentation/cassandra/2.0/cassandra/architecture/architectureSnitchDynamic_c.html http://www.datastax.com/dev/blog/dynamic-snitching-in-cassandra

Re: mixed nodes, some SSD some HD

2014-03-05 Thread Chris Burroughs
No. If you have a heterogeneous clusters you should consider adjusting the number of vnodes per physical node. On 03/04/2014 10:47 PM, Elliot Finley wrote: Using Cassandra 2.0.x If I have a 3 node cluster and 2 of the nodes use spinning drives and 1 of them uses SSD, will the majority of the

mixed nodes, some SSD some HD

2014-03-04 Thread Elliot Finley
Using Cassandra 2.0.x If I have a 3 node cluster and 2 of the nodes use spinning drives and 1 of them uses SSD, will the majority of the reads be routed to the SSD node automatically because it has faster responses? TIA, Elliot