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

[John H. Reinhardt]

On 1/1/2021 2:09 PM, Arne Vajhøj wrote:
> On 1/1/2021 11:31 AM, Dave Froble wrote:
>> On 1/1/2021 10:13 AM, Arne Vajhøj wrote:
>>> On 1/1/2021 7:17 AM, Phillip Helbig (undress to reply) wrote:
>>>> Unless they need a solution where the application can survive the loss
>>>> of a data centre.
>>>
>>> That really has not much to do with VM vs non-VM or VMS vs non-VMS.
>>>
>>> Data center redundancy is a common concept.
>>
>> Thinking a bit more about this ....
>>
>> I have no experience using VMs, other than a bit of learning a few months back.  Thus I'm not real familiar with moving VM instances.
>>
>> But what about a situation where flood waters are rising at one data center?  How much better would it be to be able to move VM instances to an alternate data center vs just shutting down the local data center and letting existing VMS instances at alternate data centers continue to run?
>>
>> Seems to me to not be much of an advantage, but what do I know?
> 
> I believe VMWare's ability to spin up a VM on a different ESXi server
> is mostly an "within data center" thing. If you have 3 ESXi servers
> and one of them get a hardware problem then VMWare can move the VM"s
> to the remaining two.
> 
> In case of a total data center meltdown, then I believe something
> else is needed. But even though OS clustering is still somewhat rare
> then application clustering is very common.
> 
> Arne
> 

While not a DR thing, the company I work for has moved data centers twice in "recent" times (once since I started 3 years ago) by just installing VMWare hosts in the new datacenter and migrating the guest VM to the new place via networks.  It wasn't instantaneous, so in the case of flood waters rising you would need a few hours notice, but moves were made from Milwaukee to Salt Lake City and later on SLC to Plano, Texas in less then 8 hours for 30 some virtual machines.  Yes, setup is required also so it would have to be a pre-planned contingency apposed to a "Oh, crap we gotta move someplace NOW!" situation.  But no hardware was physically moved.  In one case it was timed for when the old hardware was going off lease so incorporated a hardware upgrade at the same time.

We have also done the hardware failure scenario within a datacenter and, in that case, the failover time is milliseconds and Oracle (and therefore our users/customers), in our case never knew the VM moved hosts.  For performance reasons VShpere can also move VMs around to lesser loaded hosts as required.  Also without Oracle or the users ever knowing.

-- 
John H. Reinhardt