[Info-vax] VMS and the Internet of Things (IoT)

[Stephen Hoffman]

On 2016-09-12 14:32:12 +0000, Kerry Main said:

> Over the next decade, the long term strategy should not be X86-64 or 
> PowerX or ARM, but rather X86-64 + ARM + Power9 (as market growth 
> determines). Each of these HW platforms addresses a different niche in 
> the market.

Spending time on Power will be one of the better ways to ensure the end 
of OpenVMS, IMO.

> Post 2025, one would have to assume Alpha and IA64 will be distant memories.

Those two architectures are already distant memories for most folks in 
IT.   For those few folks that remember those architectures, or have 
even heard of them.   That all given we haven't seen Kittson yet, too.

> It's been discussed extensively in c.o.v. about not putting all your 
> eggs in one basket - including your reference to what Google is doing. 
> So why would VSI deciding to not put all their eggs in one basket be a 
> bad thing?  Providing Customers with choices is seldom a bad strategy.

That "one basket" being Intel and AMD and x86-64, and betting against 
the architecture that's central to the Windows hegemony, and the 
majority of Linux and BSD servers in existence?

But then providing customers with choices can be a poor strategy — 
particularly when there's little benefit to end-users and more than a 
little cost to provide the choice for both the vendor and the vendor's 
partners.   Adding a port to Power gets you little you didn't have with 
x86-64, except for fragmenting your own OpenVMS installed base and 
ISVs, and tying up VSI and ISV development for another five years, and 
higher-priced and rare boxes, among other "advantages".

This when the OpenVMS platform desperately needs to get dragged forward 
with every iota of time and effort and focus that VSI can manage to 
expend on behalf of that endeavor.  Now.

Assuming OpenVMS is around in the late 2020s and assuming it is still 
actively developed and assuming the platform owners see the 
opportunity, then whatever platform is eating into Intel's revenues 
(and those of AMD) will be the obvious porting target.   If that's 
Power, so be it.   But I seriously doubt it'll be Power, and I 
seriously doubt Power will bring much of anything that Intel doesn't 
already offer, outside of a small slice of very high-end customers, and 
those folks with control over most or all of their own software stack.

OpenVMS is not now and won't soon be playing in HPC or other high-end 
computing, either.   Not without a whole lot of infrastructure work 
first, and of which another port will serve to delay or derail.   
Adding another architecture just adds a whole pile of development and 
support work for VSI and ISVs including a whole 'nother set of 
compilers and linker and executable-related tools work, and delays and 
fragments other and more critical development work, after all.

> Partnering with IBM is seldom a bad strategy - especially with all 
> their SW which is imho, the hidden big driver.

Partnering with IBM, in direct competition with Intel and AMD, and in 
competition with multiple ARM vendors and fabs (including Intel and 
AMD), and fragmenting what's left of the OpenVMS market?  In all 
seriousness, for what?

> In addition, most mainframe types will not likely jump on any OS 
> platform running on X86-64. Should Power9 be successful, then that 
> platform would be viewed more favorably as a mainframe alternative.

The mainframe market worked out well for DEC, of course.   And a 
bespoke high-end processor or two worked for OpenVMS for some years — 
in the form of Alpha and Itanium — but all that eventually ran up 
against the cost advantages of x86-64.   Which is right where Power is 
now...

If I were guessing, AArch64 is the eventual next target, as that creeps 
up into the x86-64 performance range, assuming the ARM vendors get SBSA 
and other details sorted.   But that's still a port that's way outside 
of any reasonable window of planning and discussion for OpenVMS.


-- 
Pure Personal Opinion | HoffmanLabs LLC