[Info-vax] New VSI Community license PAK

[Chris Scheers]

I think everyone is over thinking this.

 From what has been said, it seems that this is NOT an x86 issue.

It is a VMS 9.2 issue.

For a 9.2 node to cluster with a 8.4 node, the 8.4 node will need a 
patch kit to update the cluster code.

This is irrespective of the underlying hardware.

Please correct me if I am misunderstanding this.

What I am waiting to see is for all SCS traffic to be encrypted.  In 
this case, the encryption certificate can replace the cluster password. 
  That is, if you have the certificate, you can join the cluster.


Dave Froble wrote:
> On 10/20/2020 1:37 PM, Phillip Helbig (undress to reply) wrote:
>> In article <rmn6v2$sci$1 at dont-email.me>, Dave Froble
>> <davef at tsoft-inc.com> writes:
>>
>>>> Maybe the EV6 requirement was for the version of Alpha which will
>>>> cluster with x86.
>>>>
>>>
>>> That doesn't make any sense.
>>>
>>> At the level of clustering the individual system architecture should not
>>> matter.  Just the capabilities of the software.
>>
>> I need to come up with a migration path from mixed EV56/EV6 cluster with
>> VMS 7.3-2 and 8.4 to VMS auf x86.  It can't be done in one step.
> 
> When VMS Cluster is available on x86, it is my understanding that you 
> will be able to form a cluster with both HW types.
> 
> If the cluster software on x86 is different, and I seem to recall 
> reading it will be different, then an Alpha will only be able to form a 
> cluster with x86 if the Alpha is running whatever version of VMS that 
> VSI would release in the future to allow such a cluster.  That Alpha VMS 
> release doesn't exist right now, at least in the wild.
> 
>> The basic idea is to build a mixed-architecture cluster with HBVS
>> members on nodes of both types, then gradually move to shadowset members
>> only on x86 nodes then move to only x86 nodes.
> 
> When VSI is ready for this to happen, it should be doable.
> 
>> I can't cluster what I have now with x86 VMS.  I at least have to go to
>> VSI VMS.  But I'm pretty sure that I have to go to EV6 for that.
> 
> Not saying anything definite about this thought, but, right now, as I 
> understand it, the source code for VSI Alpha VMS V8.4 2L1 and VSI Alpha 
> VMS V8.4 2L2 is exactly the same.  No differences.  2L2 was compiled 
> with compiler switches to use EV6 instructions, which I seem to recall 
> Clair mentioned that it had a performance increase just in the OS of 
> around 15%.  So a version without the EV6 only instructions should run 
> on any "supported" Alpha HW.  Not getting into anything like the Multia.
> 
> So, I don't have any idea where you got that EV6 only concept ....
> 
>>  So
>> maybe it's not clustering which requires an architecture, but rather
>> clustering (in the special case of mixed versions and mixed
>> architectures) requiring a version of VMS which in turn requires some
>> minimum hardware version.
> 
> VSI is sort of busy, so, I don't see them taking any time to remove any 
> support for older HW.
> 
>> Even if I could plug my SBB disks into x86, I wouldn't do it, since I
>> still need a mixed-architecture cluster to compile and compare before
>> switching off the ALPHAs.  (This is similar to how I went from VAX to
>> Alpha.)
> 
> You're going to want new storage on x86, perhaps SSD, or a mixture of 
> SSD and HHD.
> 


-- 
-----------------------------------------------------------------------
Chris Scheers, Applied Synergy, Inc.

Voice: 817-237-3360            Internet: chris at applied-synergy.com
   Fax: 817-237-3074