[Info-vax] VSI strategy for OpenVMS

[kemain.nospam at gmail.com]

>-----Original Message-----
>From: Info-vax <info-vax-bounces at rbnsn.com> On Behalf Of Arne Vajhøj via
>Info-vax
>Sent: September-14-21 7:25 PM
>To: info-vax at rbnsn.com
>Cc: Arne Vajhøj <arne at vajhoej.dk>
>Subject: Re: [Info-vax] VSI strategy for OpenVMS
>
>On 9/14/2021 5:04 PM, Phillip Helbig (undress to reply) wrote:
>> In article <shq41f$1ca5$1 at gioia.aioe.org>,
>> =?UTF-8?Q?Arne_Vajh=c3=b8j?= <arne at vajhoej.dk> writes:
>>>>>> There are certainly many big VMS customers who run clusters partly
>>>>>> for increased computing power.  Of course, disaster tolerance,
>>>>>> rolling upgrades, and so on are also reasons.
>>>>>
>>>>> 30 years ago: certainly.
>>>>>
>>>>> Today: not so sure.
>>>>>
>>>>> A single Itanium got lots of power.
>>>>
>>>> Sure, but what are big VMS customers doing today?
>>>
>>> Most likely running the same applications they did 30 years ago.
>>
>> Are you sure?  Some are running SIMILAR applications, which however
>> have grown so that orders of magnitude more power is required.
>
>The investment in VMS applications seems to have been relative low for the
>last couple of decades.
>
>The large new features seems to have been implemented outside of VMS
>most places.
>
>The VMS applications has been maintained and enhanced as business
>requirements has evolved.
>
>But way more +5% code projects than x5 code projects.
>
>>>>> Disaster tolerance can be achieved in different ways.
>>>>
>>>> But none as good as a VMS cluster.
>>>
>>> Businesses are interested in application availability. They do not
>>> care whether that is provided by OS features or application features.
>>
>> They should care if they have to hire twice as many people to
>> implement the non-VMS solution.
>
>But why should they need to hitre more people??
>
>Let us take Rdb vs Oracle DB aka Classic with RAC (real RAC not RAC One).
Both
>use a DLM but Rdb use VMS DLM while Oracle DB RAC use its own DLM.
>
>Does it matter for developers or operations? Not really.
>
>Arne
>

While service availability is what concerns end users and/or the business
types, in today's typical world, requiring every application developer to
understand all of the nuances of availability, multi-site site clustering,
add/removal of servers, data partitioning, data integrity is a massive
amount of inefficiencies. 

As an example, read this article about what App developers need to be
concerned about when they are in charge of addressing HA, data integrity,
workload distribution and using typical shared nothing OS strategies:
<http://highscalability.com/blog/2015/10/12/making-the-case-for-building-sca
lable-stateful-services-in-t.html>

Imho, App developers should be focussed on optimizing their applications to
meet business functionality requirements (there are typically large queues
of new functionality requests) and let the other concerns be handled at the
DB and OS / network layers.  

In terms of server clustering, there is only 2 cluster types - shared disk
and shared nothing. Depending on Application designs, there are pro's and
con's with each strategy. The typical OpenVMS cluster is an example of a
shared disk cluster strategy. 

Reference:
<http://www.benstopford.com/2009/11/24/understanding-the-shared-nothing-arch
itecture/>
"Shared Disk Architectures are write-limited where multiple writer nodes
must coordinate their locks around the cluster. Shared Nothing Architectures
are write limited where writes span multiple partitions necessitating a
distributed two-phase commit."


Regards,

Kerry Main
Kerry dot main at starkgaming dot com





-- 
This email has been checked for viruses by AVG.
https://www.avg.com