Lifespan for SSD in VMWare environment

[andong]

Hi Folks,

I am setting up a lab server using HPE DL380p Gen10, SSD generally has great performance improvement so it has became our standard configuration for the server. However different type of SSD from HPE, including Read Intensive (RI), Mixed Use (MU), Write Intensive (WI) and recent Very Read Optimized (VRO) has quite different price as well as the lifespan.

Our supplier recommended us the VRO type for cost effective purpose. for 1.92TB SSD the worst case its lifespan writes is 700TB, while reading it's specification it seems the lifespan is impacted by block size and write mode (random or sequential), apparently larger block sizes and sequential writes provide greater endurance, see figure below:

so here are my questions:

• In VMware level such as VMFS, how do I enforce the larger block size? and how do I know if the written mode is random or sequential?
• In RAID level, I am considering RAID 5, RAID 6 and RAID 10, they have different read/write gain, is there any impact on SSD endurance?
• In VM level, is there any storage policy or configuration which can be optimized in order to provide better endurance?

I am planing to install latest vSphere 7 on the host, any advice will be highly appreciated. Also it will be great if anyone could share your experience with HPE VRO SSD.

 

 

[Duncan Epping]

You can't force a specific block size with VMFS. The blocksize you are referring too is the blocksize of the filesystem within the Guest, so for instance the Windows file system. 

RAID will have an impact on lifespan, for instance with RAID-5 and RAID-6 there are parity blocks that need to be calculated and rewritten. Than again. With RAID-5 and RAID-6 you also require less storage capacity in general, so that may compensate for the parity writes

And lastly, no there's no such a thing as a policy optimized for endurance.

[andong]

Hello Depping,

Much appreciated for sharing the insights. so block size written to SSD will be controlled in guest OS level, it seems like there are some software which could enforce sequential writing rather than random writing, this may help with the SSD lifespan?

In VMFS level, the blocksize is 1M, so does it mean the vSphere writes to physical RAID in 1MB block?

Also appreciate your thoughts on RAID configuration, as we need to configure RAID for disk protection so some capacity overhead needs to be considered, guess those parity writes can't be avoided.

Thanks again and appreciate any further advice!

[Duncan Epping]

the 1MB block means that that is the allocation unit for vSphere, not the block size written to disk by the OS. If the OS does a 4KB write, vSphere will also do a 4K write within a 1MB allocation.

[andong]

Hi Depping,

Much appreciate for the clarification!

Kind Regards,

Andong