[Info-vax] Time to turn DECUServe into a mixed VMScluster?

[Arne Vajhøj]

On 6/2/2023 1:36 PM, Simon Clubley wrote:
> On 2023-06-02, abrsvc <dansabrservices at yahoo.com> wrote:
>> I have real VAX, Alpha and Itanium systems.  Perhaps if I spin up
>> an X86 version and cluster all of them together that would
>> suffice? I can add some emulated systems to the mix for fun too.
>> 
>> What would make this a PR win?   What proof would you want for
>> this? I am willing to put it together if it makes sense.
>> 
> 
> To be honest Dan, I am not really seeing the PR merits of just
> linking a current architecture (x86-64), two obsolete architectures
> (VAX and Alpha), and one mostly obsolete architecture (Itanium)
> together in a cluster.
> 
> The general reaction to that outside of the VMS world is likely to
> be "so what ?".
> 
> I should also point out you can build a Linux cluster across a
> larger range of architectures. Even if the Linux clustering
> capabilities are not as advanced as VMS, these days they can be made
> to be suitable for a good range of tasks.

You are missing the point.

The interesting aspect of the different hardware architectures for VMS
is not the different instruction sets per se but the age they belong to.

I am sure that RHEL 9 / x86-64 and RHEL 9 / ARM64 are very compatible
and interoperable.

But the VMS equivalent of that would be that VMS 9 / x86-64 and
the *non-existing* VMS 9 / Itanium are very compatible
and interoperable.

The fact that VMS / VAX, VMS / Alpha, VMS / Itanium and VMS / x86-64
are compatible and interoperable should be compared to:

SunOS 4.1 - Solaris 2.7 - RHEL 2 - RHEL 9
WinNT 3.1 - Win 2000 - Win 2008R2 - Win 2022

or something like that.

Is that a fair comparison? Not really. But marketing is
not about being fair!

> What _would_ be PR worthy is exactly what HP/HPE did 15 years ago,
> when they physically destroyed cluster nodes and then everyone
> watched while the remaining nodes recovered without data loss,
> especially if VMS was still the way fastest to recover as it was back
> then.
> 
> However, expectations have moved on since then and such an explosion 
> and simulated workload would have to be on a far larger scale than
> it was back then to be really newsworthy.

Everybody can do stuff like that today.

The fact that the functionality is typical in the applications
and not the OS does not matter much.

Arne