Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Alpha and AlphaVM benchmarks

90 views
Skip to first unread message

Artem Alimarin

unread,
May 22, 2013, 1:52:44 PM5/22/13
to
Our site (emuvm.com/benchmarks.php) has now published some CPU benchmarks for real Alpha's and AlphaVM.

You are welcome to contribute benchmarks measured on real Alpha's. The fastest EV6 and EV7 CPUs are especially interesting, because AlphaVM is in their range.

IO benchmarks are still missing. What is the best tool to measure the disk IO performance on OpenVMS/Alpha?

Keith Parris

unread,
May 22, 2013, 6:50:47 PM5/22/13
to
On 5/22/2013 11:52 AM, Artem Alimarin wrote:
> IO benchmarks are still missing. What is the best tool to measure the disk IO performance on OpenVMS/Alpha?

I like the TEST command of Stuart Rance's DISKBLOCK tool from the VMS
Freeware. It allows control of test duration, I/O access pattern (random
vs. sequential), I/O size, queue length, read/write ratio, etc. and
reports I/O rate, data rate, and response time. It can run on a raw disk
or confine its testing to within a container file on a disk to avoid
overwriting data on the disk.

Artem Alimarin

unread,
May 23, 2013, 3:56:55 PM5/23/13
to
Thanks for the advice. I have run the benchmark on a DS10 and on AlphaVM. The details are here: http://www.emuvm.com/benchmarks.php#tag_diskblock.

This test was run with image file caching disabled. With the caching enabled the history of the disk image access causes difficulties with benchmarking. The performance can increase enormously in this case. The disk image was 2GB, so it could entirely fit in the file cache of the host system.

For caching enabled, each test should probably be run several times until the result stabilizes and converges to some value. This value can than be reported as the performance with caching enabled.
0 new messages