Betreff: MSA1050 low performance with 3rd party SSD drives

[jtejada]

Hi all,

I am strugling with a case from a customer who purchased a MSA1050 with SAN controllers to be a Datastore for a VMWare7 server.  Server is a Dell R630 with Intel X520-DA2 SFP+ network card and MSA is connected directly with SFP+ DAC cables allowing 10Gbps bandwidth but they have done some performance tests and read/write IOPS are too less than expected (less than 1.000). 

The customer installed Samsung 3rd party SSD drives,  could it be that HPE limits the performance of the MSA due to the use of third-party disks?

Regards,

[Cali]

See here:

SPOCK (hpe.com)

iSCSI Initiator Notes
1) All standard ProLiant NICs are supported in conjunction with the OS iSCSI Initiator. Direct Connect is not supported.

It's not an MSA Problem, because with Windows direct connect is supported.

Cali

[jtejada]

Finally customer returned the MSA due to performance issues. I am now testing it with one of my lab Gen9 servers. This is my scenario:

DL380 Gen9 - Intel(R) Xeon(R) CPU E5-2640 v3 @ 2.60GHz - 192GB RAM

intel X520-DA2 SFP+ Card

10GbE DAC Cable directly connected to one of the MSA1050 ports

MSA with 1 Raid5 pool formed by 4 SanDisk HP DOPE1920S5xnNMRI 1.92TB SSD SAS 2.5" 6Gb drives. 

I am doing a test with FIO tool under ubuntu and I think I am also having really low performance comparing to the quickspecs table where IOPs are more than 30.000. Maybe I am doing something wrong or maybe quickspecs values are totally out of reality.

root@ERASERLAB:/lun_mount_point# fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 --name=test --filename=random_read.fio --bs=8k --iodepth=1 --size=256K --readwrite=randwrite
test: (g=0): rw=randwrite, bs=(R) 8192B-8192B, (W) 8192B-8192B, (T) 8192B-8192B, ioengine=libaio, iodepth=1
fio-3.28
Starting 1 process

test: (groupid=0, jobs=1): err= 0: pid=64490: Wed Apr 5 17:12:26 2023
write: IOPS=1523, BW=11.9MiB/s (12.5MB/s)(256KiB/21msec); 0 zone resets
cpu : usr=0.00%, sys=5.00%, ctx=32, majf=0, minf=8

root@ERASERLAB:/lun_mount_point# fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 --name=test --filename=random_read.fio --bs=8k --iodepth=1 --size=256K --readwrite=randread
test: (g=0): rw=randread, bs=(R) 8192B-8192B, (W) 8192B-8192B, (T) 8192B-8192B, ioengine=libaio, iodepth=1
fio-3.28
Starting 1 process

test: (groupid=0, jobs=1): err= 0: pid=64533: Wed Apr 5 17:12:30 2023
read: IOPS=2666, BW=20.8MiB/s (21.8MB/s)(256KiB/12msec)
cpu : usr=18.18%, sys=0.00%, ctx=32, majf=0, minf=11

[JonPaul]

@jtejada 
First you should not be using 3PAR drives in an MSA.  I'm a little surprised you were able to get them to work at all.
Next your IO test has an IODEPTH=1.  Set that to 64 or 128 and see if performance improved.
One other thing to be aware of is that first time writes to an SSD can be limited in performance.  It's typical to 'condition' SSDs when running performance.  Conditioning would run the workload for enough time that the entire SSD capacity would be written at least once.