SERT 2.0.6 Workload Efficiency score (Storage)

[Takumi Takahashi]

In SERT 2.0.6, the storage score exceeded 300 when measured with the following OS and CPU combinations.

Is this result reasonable?

Please let me know if there is any missing information.

Windows Server 2019 / i3 -12100TE
Windows Server 2019 / Celeron G6900TE
Windows Server 2022 / Pentium Gold G7400TE

Best regards,

[pgalizia]

Hello Takahashi-san,

Your Storage worklet score will vary, depending on the number of drives used, the types of drives (HDD, SDD, M.2, etc), and where they're located on the bus.

Also, be sure to disable your disk write caches as per section 4.1.2 of the SERT User Guide.

[Takumi Takahashi]

Thank you for your reply.

I checked and found that the disk write cache was disabled.

Please let me know if there are any other items I should check.

Also, is this storage score within the normal range? We would appreciate it if you could share any data that would be helpful.

[pgalizia]

Hello Takahashi-san,

You will need to run the Storage test with the disk write cache disabled (as per the SERT Run Rules section 2.4.5).  So that's correct.

As far as providing a score, we don't provide feedback on whether a score is in a normal range or not, since there are many variables involved, and any one of them can move the score one way or the other.  However, many vendors do publish their SERT scores, so you can look for other systems that are similarly configured to yours, and see if their scores appear similar.

[Takumi Takahashi]

Excuse me for continuing.

I changed the SSD capacity from 480GB to 240GB and the storage score changed significantly from about 60 to 300.

Other configurations were not changed. Is it possible to fluctuate so greatly?

[pgalizia]

No problem!

I know from other benchmarks that it doesn't take many changes to SSD configurations to cause a change in performance.  Some things you can check:

• Did just the drive capacity change?
• If this is a different drive, how do the hardware specifications compare to the original drive?  Different PCIe generations can and will perform differently, for example, or there could be a difference in firmware, even with the same manufacturer.
• Was the original drive well-conditioned, and the new drive fairly unused?  There's good literature online about conditioning SSD drives so they give consistent results when benchmarking (as opposed to a brand-new drive that can give wildly-different results until conditioning is achieved).
• Was the new drive placed in the exact same bay -- for example -- as the original?  Some systems do have different performance numbers based on what PCIe slot/bay is used for the drive.
• Did the number of drives change?  Adding or removing the number of drives and their types will cause score changes.
• And of course, were write caches disabled on the new drive?

[Takumi Takahashi]

Sorry for the late reply. I will confirm the above 6 points.

Best regards,