[Info-vax] Distributed and Clustered App Design (was: Re: WHY IS VSI REQUIRING A HYPERVISOR FOR X86 OPENVMS?)

[Jan-Erik Söderholm]

Den 2021-01-05 kl. 14:31, skrev Simon Clubley:
> On 2021-01-04, Stephen Hoffman <seaohveh at hoffmanlabs.invalid> wrote:
>> On 2021-01-04 18:17:29 +0000, Simon Clubley said:
>>
>>> So it's the same as always in that you also need the additional
>>> supporting infrastructure to provide what you get as standard with VMS
>>> clustering, including being able to detect when a node has only
>>> temporarily failed (in order to prevent everything that comes with that
>>> situation) when that temporarily failed node comes back online.
>>
>> I did a presentation on this topic way back 'round Y2K at a DECUS symposium.
>>
>> It's rather more involved than the lock manager?electing a primary is a
>> piece of this, certainly.
>>
> 
> I know about at least some of the alternatives, especially ZooKeeper
> and the components that use it.
> 
> My point was that I was surprised to see Jan-Erik suggest that a SAN
> distributed over multiple sites could by itself be an alternative to
> a VMS cluster...

Depends on *why* you use and run a VMS-cluster. Could just be of old
habits and that no one has looked at the systems since a long time.

But of course, if you use some of the unique features, it is harder
to just replace it.

> because I couldn't see how that could be possible without
> all the other bits that VMS clustering provides and which are required.

Yes, a VMS-cluster needs a lot of bits and pieces just to run in itself.
That does not mean that all application today needs the same features.
Not even an application that "used" a cluster 20-30 years ago.

> 
> Simon.
>