[Info-vax] Current limitations os VMS X86?

[Arne Vajhøj]

On 12/9/2023 9:09 AM, Robert A. Brooks wrote:
> On 12/9/2023 4:47 AM, Stefan Möding wrote:
>> Hans Bachner <hans at bachner.priv.at> writes:
>>> My latest informations says, that you can't use a shared SCSI disk as
>>> a system disk when running under VMware.
>>
>> That's also what I understand from the installation manual.  The
>> clustering manual is also only referring to Alpha or Itanium.
>>
>>> I think you *can* use a fibre channel disk as a shared system disk.
>>
>> And this would be on bare metal or also on VMware (using a raw device
>> instead of a VMDK)?
> 
> The only supported method to access fibre channel disks is if they
> are served via MSCP.
> 
> Shared system disks are not supported on X86_64.
> 
> We are investigating ESXi's fibre channel passthrough mechanism, which 
> would
> allow direct access to fibre channel storage.  Booting, however, is a 
> bit more
> complicated, so if we do support this passthrough mechanism, the initial 
> support
> would be data-only.
> 
> We realize there are ways to get booting through fibre channel 
> passthrough to work, but it takes some effort getting the relevant boot 
> drivers in place.

You need some exciting new features to put in version 10.0!

:-) :-) :-)

I suspect the demand for VMS cluster is less now than it
was 30 years ago.

In the old days N MicroVAX was cheaper than a 6000 with N CPU and
N DS series was cheaper than a GS with N CPU. But today
a x86-64 VM with 16 or 32 VCPU is nothing. So no need for clustering
to scale horizontally for existing VMS applications.

Then we have redundancy. But a lot of newer applications does
not require an OS cluster and are fine doing application clustering
over multiple standalone nodes. There are still some applications
build on VMS cluster - Rdb comes to my mind - but having 2 separate
system disks are just a minor annoyance (*).

*) And I have done it. It has always been necessary for mixed
architecture cluster (VAX & Alpha in my case).

Arne