[Info-vax] What does VMS get used for, these days?

[John Dallman]

In article <tk0r9u$1f47k$1 at dont-email.me>, news at cct-net.co.uk (Chris
Townley) wrote:
> On 03/11/2022 13:26, Simon Clubley wrote:
> > It took a quarter of a century for clustering on other operating 
> > systems to catch up to VMS clustering.
> Has it actually caught up?

Windows has not: I checked yesterday.
<https://learn.microsoft.com/en-us/windows-server/failover-clustering/fail
over-clustering-overview>. 

Back in 1997, when my employers were phasing out VMS, Microsoft announced
"Wolfpack", their attempt to port VMS clustering to Windows NT. It was ...
rather limited, shall we say. 

The manager who was most pro-Windows heard the description of Wolfpack,
and like everyone else, thought that it must be some kind of joke, apart
from not being funny. 

You could cluster exactly two machines. They needed identical hardware, a
dedicated Ethernet segment for communications, and the disks that they
could both access needed to be on a SCSI bus with the two machines at
each end. There were special drivers for the Adaptec SCSI controller you
had to use at each end, because there wasn't a single bus-master. With
all of this in place, fail-over was possible in 2-5 seconds, and
transactions in progress at the time would be lost. 

It has got better since then, but not a great deal. Cluster members seem
to still need to be running exactly the same OS version and patches,
which prevents rolling upgrades while the cluster stays up. The lack of a
good batch system makes the clustering a lot less useful. 

John