This post is more than 5 years old
1 Message
0
591
October 14th, 2017 06:00
Performance bottlenecks.
Hi All,
Ive been playing with scaleio for a couple of weeks now with Xenserver as a hypervisor still trying to get better performance out of it. We have 3 nodes.
Im wondering:
We are still on a 1 gig network as our read sequential is hitting 1 gig however the writes are no where near to the reads. If I can increase the performance to at least a gig Ill get hold of some 10gig pci cards.
1) Does the entire pool only operate as fast as the slowest drive in the entire cluster ?
2) Would 1 slower network card effect performance of the entire cluster or would that server just serve less data ?
3) Would 1 slow hard disk in the cluster effect everyones performance or would the faster drives just process more operations than the slower ones.
4) Could it effect the cluster negatively by adding in a slow PC "server" with multiple drives to try increase iops.
Thanks so much.
pawelw1
306 Posts
0
October 18th, 2017 01:00
Hi Davio,
Answering all of your questions in one - we have seen many situations where a single element (i.e. slow SDS) was affecting whole PD/SP. If you are aware of any slow components in your environment the best approach would be to remove it from the cluster and/or replace with something of similar performance to the rest of the nodes.
In Xen environment often the NIC drivers can be the culprits, so make sure you are running the most up-to-date ones and see if there are no known issues with them that require some workarounds etc.
Hope that helps.
Cheers,
Pawel