[Info-vax] WHY IS VSI REQUIRING A HYPERVISOR FOR X86 OPENVMS?

[Stephen Hoffman]

On 2021-01-01 12:17:20 +0000, Phillip Helbig (undress to reply said:

> In article <memo.20210101114248.3580A at jgd.cix.co.uk>, jgd at cix.co.uk 
> (John Dallman) writes:
> 
>> In article <98286971-2cf9-4732-a390-e2fb2e467f1fo at googlegroups.com>, 
>> ultradwc at gmail.com (D W) wrote:
>> 
>>> As for hardware failures I would I think OpenVMS clustering is a far 
>>> superior solution than the VM solution.
>> 
>> Yes, in some ways. But the VM solution is familiar to most companies' 
>> IT departments, and VMS clustering is not.

Hypervisors are generally not particularly related to clustering capabilities.

Hypervisors are rather closer to Galaxy in typical OpenVMS 
consideration and usage.

Hypervisors are very popular for consolidation, certainly. Much more 
efficient use of available hardware, so long as all guests don't spike 
together past the available reserve capacity.

Clustering multiple hypervisor guests is possible, but—like clustering 
Galaxy instances—quite possibly not the best available approach. And 
clustering across hypervisors is little different from clustering 
across Galaxy hosts.

Migrating hypervisor guests is popular with some, too—clustering 
doesn't have a good match for that, and involves manually transitioning 
the host and its file system contents whether standalone or clustered.

> My guess is that almost all existing VMS customers use clustering,

Nope. Not in my experience. Most don't. There are a *lot* of standalone 
OpenVMS systems. Far fewer clusters.

Clustering priced itself out of common usage.

> and that almost all future VSI customers are current VMS customers.

Ayup.

>  Also, modern x86-64 hardware is /fast/, and a single VMS instance 
> should be able to do the work of several Alpha boxes, or a couple of 
> Itaniums.
> 
> But increased power is only one of many reasons for clustering.

Ayup. Hardware has not only gotten faster, it's gotten much more 
reliable. And much more dense. As has software. Hardware for a given 
system load has gotten cheaper, too.

Disaster tolerance is another. Rolling upgrades, too.

>> In the common situation where VMS usage has been shrinking as a 
>> proportion of a company's IT, replacing several old machines with a 
>> single VMS VM will be attractive to management.
> 
> Unless they need a solution where the application can survive the loss 
> of a data centre.

Clustering is not the only means to that goal. With OpenVMS x86-64 
hosted on AWS, S3 provides cross-region read-after-write consistency, 
for instance. Other sites use controller-level local or remote 
replication. Some other sites use remote database logging.

And for not the first time, if you want native boot, that's absolutely 
planned and something VSI has discussed, and you'll want to coordinate 
with VSI to ensure that your preferred x86-64 hardware configurations 
are tested and supported.


-- 
Pure Personal Opinion | HoffmanLabs LLC