VMS stability back in the day (was Re: NuTek Mac comes)
Jerry Weiss
jsw at ieee.org
Sat Jul 16 11:13:45 CDT 2016
On Jul 16, 2016, at 7:55 AM, Noel Chiappa <jnc at mercury.lcs.mit.edu> wrote:
>
>> From: Jonas
>
>> At the time VMS was conceived, Unix was a university product, used for
>> teaching and research, not for heavy production work.
>
> Err, not quite. In the mid-70's, the PWB system at Bell:
>
> https://en.wikipedia.org/wiki/PWB/UNIX
>
> was being used by a community of about 1K programmers doing development of
> software for various Bell commercial projects.
>
> Yes, not accounting systems, but not "teaching and research", either. And it
> was definitely production: see the uptime statistics, etc, in the BSTJ
> article that describes it.
I was involved in a department that had university research on one side and business
on the other as well in the late 70’s and 80’s. The basic science analysis
ran on PDP-11 with UNIX variants mostly Ultrix-11, Venix and some V7.
Data acquisition was RT-11/TSX+ on LSI-11’s with custom hardware,
handlers and interfaces. The business was PDP-11’s + RSX-11, then
VAXen and VMS. Both sides did programing on Fortran and C.
Separate from the license issues in that era, we generally would have
not considered using the UNIX for the business side. While we had
or could get the technical skills to do coding for applications, the
overall support depth/response from the vendors and its operational design
was not sufficient for a small operation. If the application, media or
OS crashed, we needed to recover quickly and not risk permanent
loss of more than a few minutes of transactions.
I recall more than a few crashes on the Unix side where the file system
and data recovery was not straightforward. Even on then small disk
drives that used 60-250 Mbytes, fsck’ing could take over an hour.
The academics could afford to put a grad student on sorting though
the data loss and trying to recover missing data from multiple tapes.
Software development was slower under VMS, but the overall experience was robust.
We generally chose the tool that got the job done without too many
culture wars. Before I let we had much of the research on NeXTSTEP
or OpenStep. Steve definitely delivered a tool the academics could exploit
and we did so at every layer of that product.
Jerry
More information about the cctalk
mailing list