You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@cassandra.apache.org by Daniel Bruegge <da...@googlemail.com> on 2012/01/09 15:42:35 UTC
Cassandra on heterogeneous servers
Hi,
I am currently running 4 cassandra nodes with the following hardware
- AMD Athlon 64 X2 6000+
- 8G RAM
- 750G hard disk
It shows not such a good writing performance and a really bad read
performance with sometimes also timeouts. I was wondering if it makes
sense to add 2 nodes with a different hardware (8 CPUs and more RAM)
to improve this. Or does a cassandra cluster works best with the same
hardware in every node?
Thanks & best regards
Daniel
Re: Cassandra on heterogeneous servers
Posted by aaron morton <aa...@thelastpickle.com>.
> Or does a cassandra cluster works best with the same
> hardware in every node?
Things work best if the nodes all have the same HW. If you mix HW then you need to capacity plan against a machine made up of all of the lowest performing parts.
What's happening with the poor performance ?
Poke around with http://spyced.blogspot.com/2010/01/linux-performance-basics.html
See what's happening with nodetool tpstats ?
What sort of load are the nodes under ? How much data, how much throughout ?
Have you changed any of the default setting ?
I would guess with a single physical spindle you may have an IO bottleneck.
Hope that helps.
-----------------
Aaron Morton
Freelance Developer
@aaronmorton
http://www.thelastpickle.com
On 10/01/2012, at 3:42 AM, Daniel Bruegge wrote:
> Hi,
>
> I am currently running 4 cassandra nodes with the following hardware
>
> - AMD Athlon 64 X2 6000+
> - 8G RAM
> - 750G hard disk
>
> It shows not such a good writing performance and a really bad read
> performance with sometimes also timeouts. I was wondering if it makes
> sense to add 2 nodes with a different hardware (8 CPUs and more RAM)
> to improve this. Or does a cassandra cluster works best with the same
> hardware in every node?
>
> Thanks & best regards
>
> Daniel