I am wondering if someone could confirm or speculate regarding some performance figures we are seeing from a CX4-240.
The figures seem too slow. We would like to get a feel for other experiences if possible.
Blades: RHEL5 AMD Opteron, 8GB RAM, 2x 15K SAS mirror, 2x QLogic 4Gb FC HBAs
2x Brocade 5000 FC Switches
15x 400GB 10K FC disks
15x 1TB 7.2K SATA disks
HBA's & switches at minimal workload.
Test LUNs on separate SPs (normal, not metaLUNs or Thin LUNs)
Caching disabled on Test LUNs.Ext3 filesystems on both LUNs.
LUNs & Blade OS idle when tests were performed.
Tests performed on separate blades to confirm results.
LUN1; 3x SATA RAID5
LUN2, 10x FC RAID5
Sequential r/w, LUN1 -> LUN1: avg 13MB/s
Sequential r/w, LUN2 -> LUN2: avg 27MB/s
Sequential r/w, LUN1 -> LUN2: avg 15MB/s
Sequential r/w, LUN2 -> LUN1: avg 17MB/s
Sequential r/w, Host SAS -> LUN1: avg 16MB/s
Sequential r/w, LUN1 -> Host SAS: avg 17MB/s
Sequential r/w, Host SAS -> LUN2: avg 60MB/s
Sequential r/w, LUN2 -> Host SAS: avg 82MB/s
Any comments would be most welcome.
We are in the process of running more tests (random workload, different size RAIDs)
I have same situation.
I have checked iowait using sysstat utility on RHEL, I have discovery many value than non EMC system.
And I have checked que value on powermt utility, I have discovery many queing value.
Then I have tried to single path access, SPA-0 to a HBA port.
This access performance is very well, minimum iowait and minimum queing.
I think so PowerPath for Linux has any bugs or I need any special settings???
If you can test, please check sysstat-iowait and powermt-que value
with multi-path and single-path environment.