[Info-vax] VAX Harddisk Benchmark / VAX 4000/105A SCSI performance

[Stephen Hoffman]

On 2018-08-30 06:24:50 +0000, hans.huebner at gmail.com said:

> ...does anyone happen to have a hard disk or file system benchmark for 
> VAX/VMS that they'd be willing to share?  I need to get rid of the RF36 
> drive that is in my VAX 4000/105A as it is too loud and I would like to 
> experiment with a few SCSI options.  Having a way to measure relative 
> performance would be useful.
> 
> I've been unable to find information as to the performance of the SCSI 
> subsystem in the 4000/105A, in particular in comparison to DSSI.  I 
> tend to suspect that even if the SCSI transfer rate is slower than that 
> of the DSSI bus, a solid state drive on the SCSI bus would still be 
> overall faster than the RF36.  Then again, I could be wrong, so I'm 
> looking for a way to measure (or experience reports from knowledgeable 
> people).
> 
> I can write my own benchmark, but maybe there is some benchmarking tool 
> that can generate a specific I/O mix of random and sequential accesses 
> and report the results already?

There's little performance distance between the glacial storage of DSSI 
and the glacial storage of SCSI-1, in terms of modern storage 
performance.

The DSSI SHAC controller was theoretically capable of ~4 MBps, and 
SCSI-1 of ~1.5 MBps.  HDDs were good for ~120 IOPS and not a whole lot 
of bandwidth, which is what throttled most everything back then, and 
why ginormous arrays of HDDs were common.  In this case, the VAX and 
the memory and the rest of this old box will also throttle performance. 
 Pragmatically, I wouldn't expect much of a difference, but then I'm 
well used to SSD speeds and feeds and HDDs all seem slow.

If you have any requirements for performance, transfer your environment 
and boot a VAX emulation and use an SSD.  I'm routinely getting tens of 
thousands of IOPS with SSD, and a mid-grade array provides most of a 
million IOPS and at 0.7ms latency; at a very small fraction of the 
access time of an HDD.  And emulation on any recent Windows, Linux, 
macOS or BSD system will be vastly faster than an actual hardware VAX.

Here are some DSSI disk specs: 
http://manx-docs.org/collections/mds-199909/cd1/vax/514aatib.pdf — feel 
free to scrounge for some equivalent-era SCSI storage specs for however 
you're connecting the SCSI storage to this old VAX; via the internal 
SCSI bus or via one of the various external HSD adapters.

What follows are related discussions of OpenVMS I/O performance from 
David Mathog from a number of years ago, and a benchmark tool that 
David had written.  Lots of old discussions around here, available via 
the Google Groups archives.  I've not tried the SAF download link in a 
few years, though.   That tool that was still available back then, and 
I have a copy of the download if it's gone missing.

https://groups.google.com/d/msg/comp.os.vms/4FZHjDQ1R4A/DO5xV-z-XGEJ
ftp://saf.bio.caltech.edu/pub/software/benchmarks/mybenchmark.zip


-- 
Pure Personal Opinion | HoffmanLabs LLC