Do VSI OpenVMS Drivers support Trim for SSD?

[reinhardtjh]

The subject says it all (I think).

As a hobbyist, looking to put standard commercial SSD (Samsung EVO, WD Red, etc) in my Integrity rx2660. I already have some in my AlphaServer DS10's as data disks (unbootable). Just wondering is trim is available to help extend the life of the SSD.

The rx2660's disk controller is a P600 if that makes any difference. I also have a P400 available if that is a better option.

Thanks

[pocketprobe]

According to viewtopic.php?f=39&t=8952 this is a no. I currently use SSD storage for my bare metal x86 machine, despite the lack of TRIM support.

[reinhardtjh]

According to viewtopic.php?f=39&t=8952 this is a no. I currently use SSD storage for my bare metal x86 machine, despite the lack of TRIM support.

Thank you! Mr Brooks responded to my cross posted, uh, post in C.O.V and his remark about the HPE version of trim not being a very good implementation struck my memory and I knew I had seen it somewhere else. It's about 8 posts in the thread you list.

Also in SCSI the function is "unmap", "trim" is the IDE version.

I've got SSD in my x86 VM Host as well. There, I assume the VMWare disk drivers support trim/unmap hidden to the guest VM systems. Hopefully not an optimistic assumption.

[pocketprobe]

You're most welcome.

I will make the assumption that your host is using an LSI/Avago SAS HBA/RAID controller. These controllers will let ESXi know about the underlying storage being SSD if the SSDs are either SAS or are SATA and support Deterministic Read Zero After TRIM. If your SATA SSD lacks this feature, the controller will proceed to treat the disk as a conventional hard disk and will not pass TRIM down the stack. On a disk that does not support this, attempting to read TRIM'd blocks before the controller has a chance to erase them, you can readback deleted data. Ideally for the best experience you would have a guest that supports TRIM/UNMAP, on top of a VMFS6 volume for automatic thin provisioned storage space reclamation, on top of disks that are either SAS or SATA with DRZAT.

If the guest does not support this, it will behave just like a system where one of the layers disagrees, and block unmap will not occur. The virtual disk may support it, but the operating system will not call it.