06-17-2014 02:22 PM - last edited on 03-25-2019 01:38 PM by ciscomoderator
Hi UCS Community,
One of our server owners is claiming to have slower disk I/O performance on their B440 M2 UCS blade than on their rack mount (non-UCS) servers. The OS is Windows Server 2008 R2 and the backend LUNs are on an EMC VMAX array.
At a glance, I can't anything amiss with the Service Profile's vHBAs to suggest there might be a performance issue.
What can I look at specifically to build a case for or against an issue with disk performance rooted in the UCS?
Thanks!
06-18-2014 01:19 AM
My suggestion, open a TAC case, there's really few details to go around in your description.
We'd need information about the adapter, path info etc etc.
Two things you can check:
0) Narrow it down. Are particular VMs affected, are particular LUNs affected, are particular controllers affected?
1) Check the path along your deployment (from FIs to storage) for interface errors. Bad cable can cause quite a headache :}
------
show interface count err
------
on NXOS based platforms.
2) Check queue depth on OS side. Windows keeps per LUN queue AFAIR.
M.
06-18-2014 09:56 AM
Is the rack mount and the UCS blades connected to the same SAN; what kind of SAN ? FC, NFS, Brocade, MDS ?
We have seen such situations in the past; they were mostly not UCS related. Do you see any BB-credit starvation ?
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide