[Info-vax] free shell accounts?

[Stephen Hoffman]

On 2015-01-23 12:57:22 +0000, Stan Radford said:

> Bob thanks. I don't have an ACM membership any more so I don't think I 
> could get this. I did download a bunch of doc from HP though and I have 
> books on the way so I'll read up on this. My question is not on 
> clustering although sure I'm interested in VMS clustering just to see 
> how they did it.

Skim the manuals, too.  They're pretty good.

> My question is coming from wondering how the deathrow hobbyist cluster 
> could really be a cluster from a developer's POV if the file(s) he 
> needs right now
> are physically connected to only one box and that box dies and I 
> understand that this is a problem which brings me to wonder in what 
> sense was deathrow
> actually a cluster.

Clusters are about sharing.  The reliability against failures arises as 
much from the hardware configuration and from how the software can use 
that.

Deathrow is a single security domain, and has various core system files 
and all of the user files are shared across all hosts, and the files 
are read-write shareable to the record level across all members.

> I do realize production is quite different from development and of 
> course it is much easier to plan for a stable production environment 
> and replicate what is needed so it exists in multiple copies across a 
> network, cluster etc. than it is to deal with a development environment 
> where code is being worked on regularly.

An open-access hobbyist cluster run on donated hardware is yet different.

If you're working on source code and sharing that with other 
developers, then using a DVCS would be more typical these days.  That's 
obviously quite different from clustering.  Now if you're looking to 
share your application data and to keep your application or your DVCS 
accessible over hardware failures or data center outages, then an 
OpenVMS cluster will probably be of more interest.

Clustering is a somewhat different approach and that combines factors 
you might expect from a DVCS, LDAP and NFS, as well as providing a 
distributed lock manager that can coordinate the access to arbitrary 
resources.

Full disclosure: I've written part of the VMS manuals, as well as 
several of the books on VMS.  I'm one of the Deathrow admins.  No, I 
don't know when the current routing mess will be resolved and access to 
the cluster restored.


-- 
Pure Personal Opinion | HoffmanLabs LLC