[Info-vax] "x86 has only a few years left in the market place"
Kerry Main
kemain.nospam at gmail.com
Mon May 21 16:40:44 EDT 2018
> -----Original Message-----
> From: Info-vax <info-vax-bounces at rbnsn.com> On Behalf Of Stephen
> Hoffman via Info-vax
> Sent: May 21, 2018 11:30 AM
> To: info-vax at rbnsn.com
> Cc: Stephen Hoffman <seaohveh at hoffmanlabs.invalid>
> Subject: Re: [Info-vax] "x86 has only a few years left in the market
place"
>
> On 2018-05-21 14:38:23 +0000, seasoned_geek said:
>
> > On Thursday, May 17, 2018 at 12:18:46 PM UTC-5, Stephen Hoffman
> wrote:
> >>
> >> Wasn't a whole lot of call for too much and too high-end with
> OpenVMS,
> >> based on the last decade or two...
> >
> > There was always call, it was never delivered. At it's height when
DEC
> > still was DEC and we were pushing into data centers with massive
blue
> > iron we kept failing the minimum throughput tests at the companies
> who
> > wanted the cost savings and robustness of VMS but could not/would
> not
> > sacrifice throughput.
>
> DEC at its height was a rather different era of computing, and most of
> thirty years ago. Given that DEC wasn't able to profitably build big
> iron that met those throughput requirements? The so-called VAX
> mainframe didn't do all that well in the market, after all.
>
> Any of the DEC attempts to compete for customers with IBM hardly
> seems
> relevant to VSI and OpenVMS in 2023 and 2028. That particularly
given
> that in the intervening years between DEC at its height and VSI was an
> era when the platform qualification effort for existing Superdome iron
> wasn't viewed as sufficiently profitable to warrant the effort.
>
> n.b.: There was no native HP/HPE SD1 boot support for OpenVMS.
> HP/HPE
> SD1 support for OpenVMS via HPVM and HPUX, and OpenVMS on that
> stack is
> no longer being qualified and updated. OpenVMS was never qualified
> on
> HP/HPE SD2.
>
> VSI will eventually have the option of qualification on HPE SD-X and
> ilk, if the VSI folks are inclined and if there are enough folks
> willing to purchase. Those are likely going to be some expensive
> software licenses, too.
>
> And I'm exceedingly skeptical that there'll suddenly be notable call
> for support of HPE SD-X class gear. Not the least of which is because
> processors and I/O and the rest of servers are substantially faster
> than in the DEC era.
>
> In the past decade or two, I'm not aware of a great deal of call for
> high-end servers with OpenVMS. Folks in recent years have been
> maintaining their existing OpenVMS installations. Most of which
> apparently don't want to pay for SD1- or SD2-class servers and the
> related qualification costs. Maybe we'll see OpenVMS qualified on HPE
> Apollo or some of the other similar boxes. But I'd expect that the
> vast majority of the VSI OpenVMS server market will be two-socket
> servers, and quite possibly even those part-populated. Computing
> server density has gone up slightly since the VAX era, after all.
>
One needs to remember:
1. Reason why HP-UX, Solaris, AIX Customers always wanted big
Superdome+ servers was that their only real native answer to scalability
was scale up. HP-UX, Solaris and AIX do not have very good native
clustering scale out capabilities. OpenVMS Customers can scale clusters
up AND scale out.
2. The demand for larger servers running OpenVMS was definitely there,
but pricing and overall cost was definitely a reason why many Customers
started turning off big boxes. Course, that is true for all UNIX
platforms as well. Solaris, AIX, HP-UX product teams were competing with
each other on bigger boxes while Linux (and to a lesser degree, Windows)
was eating all their lunches by providing "good enough" solutions on
X86-64. Sun, IBM, and HP beat themselves for not recognizing where the
true competition was coming from.
3. Pricing for VSI in future will likely be more similar to Linux model
i.e. monthly support subscriptions, so future overall OpenVMS costs will
likely be a fraction of what HP used to charge.
4. Often forgotten is HA when using large servers. As expensive as 1
large server was, if HA was important, then you had to buy 2 big
expensive boxes.
Regards,
Kerry Main
Kerry dot main at starkgaming dot com
More information about the Info-vax
mailing list