[Info-vax] Defrag utility needed on Integrity SSD-only configurations?

Stephen Hoffman seaohveh at hoffmanlabs.invalid
Mon Sep 9 16:06:21 EDT 2019


On 2019-09-09 18:50:48 +0000, Rod Regier said:

> ...Any thoughts on the usefulness/liability of regular execution of a 
> disk defragmentation utility on OpenVMS executing on an Integrity 
> server only
> equipped with SSD mass storage?
> ... Just for reference/comparison purpose, MS Windows 10 occasionally 
> defragments SSD's on an as-needed basis.

Main reason for defragging on HDD is the I/O head seek performance.  
Which is not a consideration with SSDs.

Ongoing defragmentation?  No.  But your environment(s) may differ than 
what I'm dealing with.

I'd look at this operation on a per-file basis, in conjunction with 
analyzing how (badly) the local crop of RMS files have been consuming 
extent headers.

Sector-level fragmentation does consume extent headers, and RMS indexed 
file formats commonly internally fragment, too.

This is also all part of the usual RMS CONVERT, and related file system 
maintenance baggage.

And you'll want to be setting the locally-appropriate RMS file extent 
sizes, same as usual.

But pragmatically, no, I wouldn't run a regular defragmentation pass.

I'd look to DFU if and as needed, and would use BACKUP /IMAGE 
disk-to-disk during maintenance windows.

As usual with BACKUP /IMAGE, the target will be defragged.  And you'll 
have a complete and consistent and quiescent backup.

Given reasonable extent sizes and non-stuffed volumes, the usual RMS 
CONVERT operations will also implicitly defrag the associated RMS files.

RMS classically just doesn't do well at the combination of perpetual 
uptime and consistent backups.

Splitting a RAID set—whether host based volume shadowing shadowset 
volume or otherwise—for a backup won't help here.

FWIW, DFU and most (all?) of the OpenVMS defraggers use the same 
OpenVMS IO$M_MOVEFILE API for defragging.

OpenVMS doesn't support TRIM, so you'll want to enable erase-on-delete 
if (only if) the SSD firmware provides support for that optimization.

BTW, here's a better write-up of what Microsoft Windows is up to here, 
and why, and with links to supporting information:

https://www.hanselman.com/blog/TheRealAndCompleteStoryDoesWindowsDefragmentYourSSD.aspx 


It'll be interesting to see how the new VSI VAFS file system will deal 
with all of this, but that's probably still a few years out.

No, I don't immediately recall the maximum number of file fragments 
permissible in ODS-2 and ODS-5.  It's pretty big.

-- 
Pure Personal Opinion | HoffmanLabs LLC 




More information about the Info-vax mailing list