[Info-vax] Marketing ideas for VSI ?

[Rich Jordan]

On Friday, December 14, 2018 at 5:44:56 PM UTC-6, Stephen Hoffman wrote:
> On 2018-12-14 16:49:46 +0000, Rich Jordan said:
> 
> > What Hoff said remains a (no longer 'the') major problem with 
> > clustering; the excessive cost.
> 
> Clustering pricing was routinely US$25K per host in various quotes, and 
> this was for low-end single-socket Alpha server boxes.
> 
> This pricing has been utterly untenable in a number of the quotes I've 
> worked, and has either wholly blown apart or has led to severely 
> down-graded configurations.
> 
> > Cluster in a box.
> > Two decent servers, shared storage or shadowing with a cluster 
> > interconnect (even if it is just GbE or 10GbE)
> 
> Either three servers, or two servers with a cluster shared storage 
> interconnect.
> 
> Two servers without a shared interconnect is a primary-secondary 
> configuration at best, and more typically a baroque and expensive 
> replicated-storage failover configuration.
> 
> 10GbE minimally.  Faster would be preferable.  And my usual reference 
> around network I/O stack performance: https://lwn.net/Articles/629155/
> 
> > VMSCluster (and maybe shadow) licenses included
> > Rdb or other relevant database license included (at least runtime)
> > Fixed price program development option including at least one compiler 
> > of choice and tools (more compilers=better); database dev license also 
> > if base is runtime.
> > Unlimited user licenses
> > TCPIP Services
> > Bundled service/support for (1 year?) with 'special' renewal pricing 
> > relative to having bought all the stuff independently.
> 
> Oracle would seem an unlikely partner in this effort, though stranger 
> things have happened.
> 
> PostgreSQL or SQLite are alternatives and can be good options, and that 
> incorporation would address other longstanding issues within the base 
> distro and its apps.  Downside: PostgreSQL is wedged behind SSIO.
> 
> The VSI debates around traditional license prices, around support-based 
> prices with low-cost or free licenses akin to Centos and Fedora and 
> with RHEL when support is desired, and around a migration to software 
> service (rental) pricing akin to iCAP or otherwise will undoubtedly 
> continue.  The OpenVMS x86-64 licensing was the usual baroque 
> collection of boxes and arrows—simplicity is seemingly an anathema to 
> software licensing—and it seems likely we'll learn more about that 
> licensing as the production release approaches.
> 
> Yes, having a pre-packaged configuration would be interesting.  BTW: 
> DEC offered this sort of thing with the VAX 6333, VAX 8974 and VAX 
> 8978, and a few other little-known VAX "models".  VAX 6333 list price 
> was between US$2M and US$3M depending on storage, BTW.  It wouldn't 
> surprise me to see a recent smartphone or tablet outrun that most VAX 
> configurations, either.  But I digress.
> 
> OpenVMS has no entry-level pricing.   But those same licenses are the 
> lifeblood of VSI.  No good answer, here.
> 
> 
> 
> -- 
> Pure Personal Opinion | HoffmanLabs LLC

Yep, pardon the brain fart; shared storage required unless the third node is cheaper than viable shared storage.  Maybe if VMS on intel can provide good iSCSI support, then a suitable quality small SAN or even NAS might do for the storage.

Right now I would expect 100GB networking would have to be an upgrade at the hoped for price; 10GbE is getting closer to being commodity priced so I'd hope would be the minimum, at least for the cluster interconnect if not the primary data connection(s).

I've wondered if DEC, or now VSI could make a minimal 'quorum node' to live on the network and do nothing but the bare minumum to be a low resource low cost (and low hindrance) 'vote' for a LAVC.  Some of our customers kept an old VAX or VAXstation around for that purpose when they downsized from VAX/11 systems with star couplers and HSCs to microVAXen or early Alpha LAVCs; they did nothing any more but act as the third node.