Hi and Happy New Year everyone,
I have a problem with read/writes to/from drive connected as iSCSI. Setup:
- host: HP Proliant DL360e Gen8; hypervisor on SD card
- 4x SSD as internal datastore. VM files (including primary disk) sit on it.
- VM's OS: WS2012 R2.
- 1x 1Gbps adapter assigned to VM.
- secondary disk: 4TB (ESXi limit) vmfs file sitting on 9TB LUN, mapped to an iSCSI target, located on QNAP (series x79, latest firmware), mapped as vd (rdm greyed out).
- iSCSI setup very simple - default options/values on ESXi host's iSCSI Software Adapter, one 1Gbps port on each side (QNAP and ESXi), CHAP enabled.
- Store on QNAP on RAID6, 10+ 4TB 7.2k drives, all healthy.
If it's not obvious from above: datastore mapped OK, over 1 path, drive mapped OK, accessible from VM. But:
Max write speed I've achieved was <30MBps; max read speed was somewhat lower (~20MBps, iirc). This itself I already would consider being an issue, since that's about a quarter of 1Gbps capacity.
Biggest issue however is, that if BOTH read and write occur at the same time, read speeds basically crawl - speeds vary from 20 to 100KBps...
I'd have few questions related to above:
1) is that normal behavior? I could understand lower performance on single sequential access drive while doing write and write at the same time, but still not this. Is it maybe something in the iSCSI design, that I'm not aware of? I must admit, I don't know all its ins and outs.
2) if that is not normal, can you see any obvious mistakes in above setup? I.e. could multipathing with binding help here - of is it rather for HA/redundancy purposes?
3) any other advices/tips you can give me? I.e., if that setup is not viable for the purpose (FTP backend), would NFS be an alternative? Or even NAS mapping? I'd like to get the iSCSI to work though - would be useful to build redundancy/backup.
Any help would be greatly appreciated.
Best,
Chris