Hello all,
we have a couple of DL360 G5 machines. All of them have the Smart Array P400i RAID controllers with 512 MB BBWC.
In one machine (let's call it "A") we have built 2 RAID arrays:
- 3x 72 GB 10k rpm SAS disks in RAID-5
- 3x 500 GB 7.2k rpm SATA disks in RAID-5
This machine has only 8 GB of RAM installed (8x 1 GB).
The other machines (let's call them "B") have only one RAID array configured: 3x 144 GB 10k rpm SAS disks in RAID-5. These other machines have 32 GB of RAM installed (8x 4 GB).
We're facing an interesting problem: the disk IO performance of machine "A" is really slow in case of both RAID arrays. When we use the faster array (with the 10k rpm SAS disks) the average speed in random write/read tests is 26 MB/sec. When we use the slower array (with the 7.2k rpm SATA disks) the average speed in random write/read tests is 13 MB/sec.
On the "B" machines we have significantly faster IO speeds: more than 450 MB/sec in average in the same read/write tests.
The tests were performed with the Phoronix Test Suite (same version and same benchmarks on all the machines, with the same settings).
I checked machine "A" with the HP Insight Diagnosis tool and the health status of all the disks and of the RAID controller is OK, there are no errors indicated. There's one non-related (at least in my opinion it shouldn't cause these problems) error: the "serial port #1" is faulty.
Is it possible that using SATA and SAS disks on the same controller causes this very slow disk IO performance? Initially I ruled out this option, because if the RAID controller would run in the "fallback mode", it should be still faster than the performance I measured. Or am I wrong?
Is it possible that the faulty "serial port #1" affects the disk IO performance?
Thank you in advance