[Info-vax] Marketing ideas for VSI ?

[Stephen Hoffman]

On 2018-12-15 08:14:34 +0000, Phillip Helbig (undress to reply said:

> This is an argument for a collection of bundled licenses.  But someone 
> wanting to pay for a cluster would want them at different sites.

Short:

A site-specific bundled cluster limitation would be new.

If somebody wants to relocate part of a packaged cluster configuration 
remotely and has network pipes to match, then there's nothing that 
would preclude that going all the way back to the VAX 6333, VAX 8974, 
VAX 8978 and ilk.


Long:

Outage costs vary widely.  As does the quality and capabilities of and 
features of app software.

The disaster-tolerant and multi-site cluster configurations of fondest 
marketing fantasies are certainly occasionally seen in the wild, though 
those configurations are a smaller subset of all clusters than many 
might realize.

More than a few folks—those that can afford the licensing—use 
clustering for faster failover from hardware failures. Usually within a 
single data center.  In a number of cases, clustering provides for 
faster restarts of apps and configurations that are effectively or 
entirely single-host, too; for what replication support now provides 
many folks.

There are a number of apps that do have some cluster awareness, though 
those are far from ubiquitous, and the installations and setups tend to 
be arcane at best, and far fewer are capable of continuous access 
across app-level rolling upgrades.  And there are often holes around 
consistent online backups, and around file maintenance tasks.  That 
apps are bad at clustering and that clustering is expensive and that 
there are no entry-level cluster offerings may well be connected, too.  
Just spitballing there, though.

At the low- to mid-range part of the OpenVMS installed base, I 
routinely encounter manual-failover configurations.  These are the 
good-enough configurations, due to software limitations or the 
stratospheric cluster license costs, and the usual trade-offs between 
outage costs and license costs and software costs.  A subset of these 
failover configurations are due to software limitations; 
non-cluster-aware or cluster-limited apps.

The documentation and the frameworks associated with designing and 
writing cluster-aware software are and as I've occasionally commented, 
atrocious.  I know how this stuff works, and it's arcane. And fussy.  
Which is a problem for robustness and continuity.  And retrofitting 
cluster awareness into an app is non-trivial.

For those that really do want or need multi-site configurations and 
that have the budgets to match their outage costs and that have bespoke 
apps and a budget to maintain and update those, I'd tend to suggest 
ordering two boxed clusters and redundant data links.




-- 
Pure Personal Opinion | HoffmanLabs LLC