I recently ran a few tests of my TS-467A device (2 x NVMe, 2 x SSD and 4 x SATA) running QTS.
Here are my observations for sequential write tests:
a/ Qtier on all drives (NVMes as ultra-high, SSDs as high and SATAs as base tier) - writes at full speed, however SSDs are bypassed/not used,
b/ Qtier on SSDs (ultra-high) and HDDs (cold tier), NVMes configured as cache - writes at full speed at the beginning, but dramatically drops on SSDs (latency reachs a few thousands of msecs, copy operation hicks-up after copying around 100GB),
c/ Qtier on NVMes and HDDs (no SSDs) - writes at full speed all the time.
I wonder if it is normal that SSDs are not used in scenario A. I tought that storage layout described in scenario A would be optimal. Is anybody facing the same?
To help us better understand and troubleshoot the issue, could you please tell us what specific method you’re using for your tests? Knowing your exact testing approach would be a great help in pinpointing the problem.
I am using simple copy operation from external source over smb protocol. One stream consisting of a couple of 20GB files.
FYI I can’t cross 115MBps, since I am using single 1Gbps connection.
Thanks Steve.
Interesting article. Perhaps I should redesign my storage and configure NVMe as cache and SSD as Ultra High Speed tier.
Would it be helpful if I raise a support ticket?
Yes, if you have any other questions, would like our internal team to assist with further confirmation, or have logs you can provide us, please feel free to open a support ticket with our Support team. They can perform a detailed check.
Of course, the forum is an open discussion space, and the benefit here is that other helpful members might be able to assist and discuss the issue with you as well.