[Info-vax] HP Integrity rx2800 i4 (2.53GHz/32.0MB) :: PAKs won't load

[Kerry Main]

> -----Original Message-----
> From: Info-vax [mailto:info-vax-bounces at info-vax.com] On Behalf Of
> Stephen Hoffman via Info-vax
> Sent: 28-Feb-16 10:29 AM
> To: info-vax at info-vax.com
> Cc: Stephen Hoffman <seaohveh at hoffmanlabs.invalid>
> Subject: Re: [New Info-vax] HP Integrity rx2800 i4 (2.53GHz/32.0MB) ::
> PAKs won't load
> 
> On 2016-02-28 14:20:06 +0000, Kerry Main said:
> 
> > I agree with Hoff that simplifying OpenVMS mgmt. functions like
> > clustering,multiple data repositories, upgrading security etc. should
> > be considered forfuture next gen versions of OpenVMS.
> > Having stated this, to the point raised in the last reply, one needs
> > tounderstand that multi-site clustering with no data loss in a DR
> > scenario isa really tough nut to crack - on ANY OS platform.
> 
> Unnecessarily complex, baroque, arcane, variously idiotic and
> idiosyncratic, inconsistent, difficult to maintain, difficult for the
> system and application developers and the to extend, difficult to
> document, and difficult to troubleshoot.  ~Thirty years of designs that
> were good at the time, and various designs that were expedient, and
> now
> buried under decades of accretions and incidentals — and the ubiquitous
> defense against all substantial OpenVMS enhancements known as
> "compatibility", of course.
> 
> The current cluster user interface "design" is utter crap.    Nobody
> would use this "design" today.   Nobody.   Yet here we are.
> 
> Modern clustering is increasingly a solved problem, too.   With data
> replication, multiple sites and the rest all available.   Increasingly
> at scales vastly larger than the OpenVMS 96 host limit.    Solutions
> that are at least suitable for the needs of most customers, and with a
> budget that they are willing to pay.   This is the market that VSI is
> playing in now, too.
> 

I know you know there is a difference between replication, data 
consistency, and not losing data in a significant event, so I won't digress 
too much here. 

For others not familiar with Cloud computing, cloud providers have often 
promoted data replication as their answer to not doing backups or inter
site consistency SLA's.

> These competing solutions are not using the designs — and variously,
> the "designs" — that DEC had envisioned.  The world is clearly getting
> by, and with very little use of OpenVMS and clustering.    This is
> where current and potential future customers and ISVs are operating,
> and is what current customers and ISVs are comparing OpenVMS with.
> 

No matter what fancy cluster or multi-server designs are out these today
on ANY OS platform, they all boil down to combinations of active-active, active
-passive using shared nothing or shared everything designs.

Each of these basic designs has pro's and con's with associated trade-offs.

Here is good whitepaper on shared-nothing vs. shared everything:
http://tinyurl.com/data-sync-challenges  

original (will wrap)
http://www.scaledb.com/wp-content/uploads/2015/11/Shared-Nohing-vs-Shared-Disk-WP_SDvSN.pdf 

And in case there is any doubt about the significant challenges out there
today with existing *nix /commodity OS architecture deployments, here
are some good links that state otherwise i.e. state-less vs stateful arch's
from a former Microsoft Game developer and now a distributed systems
engineer at Twitter: (links Hoff supplied awhile back - *really* good links)

"Building Scalable Stateful Services" by Caitie McCaffrey
https://www.youtube.com/watch?v=H0i_bXKwujQ

Making the Case for Building Scalable Stateful Services in the Modern Era
http://tinyurl.com/stateless-stateful  (references Catie's talk above)

> I've had more than a few existing OpenVMS sites look at cluster
> projects and get nuked in the pricing stage, too.   Existing OpenVMS
> sites that looked at clustering prices and/or clustering complexity
> and/or the code changes and decided they'd get by without it.   But I
> digress.
> 

I agree cluster and OpenVMS pricing / licensing  in general needs a major 
re-visit for next gen OpenVMS V9/10+ futures ..

The competition in the future and has been for the last 10+ years is not 
Solaris, AIX, z/OS etc, (DEC/Compaq/HP market positioning) but rather 
commodity OS's like Linux/Windows. This is especially true when VMS is
being sold on the same server X86-64 architecture as Windows/Linux.

> VSI can do vastly better here.   VSI will want to do better here.
> Particularly because user interfaces based on half-baked and
> thrown-together designs with no clear and no overarching model — and
> the current gonzo-scale prices — just aren't going to attract much in
> the way of new customers and new developers and ISVs.   Or in various
> cases, these won't even existing customers and partners.
> 
> We are no longer in the era of servers as pets.   OpenVMS and any other
> server needs to operate as livestock.  In herds.  Copying config files
> back to the system disk to do an OS upgrade is a non-starter.
> Isolation of authentication and logging is a non-starter.   Baroque UI
> designs.   Etc.
> 
> But again, I'm being polite.
> 

I don't believe anyone here is arguing that major licensing, pricing, technical, 
security and yes, simplification changes, are not required at some point in the 
future. 

However, like all companies, there is only so much one can do with existing 
resources and budgets, so trade-offs and delivery schedules need to have a 
balanced perspective.

Regards,

Kerry Main
Kerry dot main at starkgaming dot com