Re: Could a PC do this?
From: Jake Hamby (jhamby)
Date: 05/11/05
- Next message: David J Dachtera: "Re: Clusterwide QUEMGR file"
- Previous message: Wayne: "Letters to the new CEO"
- In reply to: John Smith: "Re: Could a PC do this?"
- Next in thread: Dave Froble: "Re: Could a PC do this?"
- Reply: Dave Froble: "Re: Could a PC do this?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Tue, 10 May 2005 18:04:03 -0700
On Tue, 10 May 2005 10:31:00 -0400, "John Smith" <a@nonymous.com>
wrote:
>We look at Itanic's less than stellar performance both in sales/adoption
>rate and outright performance and cringe.
I'm a bit surprised that people seem to be expecting a totally new
architecture to immediately be competitive with x86. Perhaps it's
Intel and HP's fault for overhyping the platform, combined with the
general impatience and short-sightedness of the industry, but I
wouldn't expect Itanium to be competitive *until* 2006 or 2007 and
then I would expect its performance to gradually surpass what can be
done with x86-64, particularly for floating-point (where x86 has
always been weak) and in 16-CPU and above configurations.
As for sales/adoption rate, I just read about SGI's new Prism system
which is their new super-high-performance Linux workstation for OpenGL
applications. Now that SGI has abandoned MIPS and gone the PC route,
why is Prism (as well as their Altix mid-range servers) based around
Itanium 2 instead of Opteron? Is SGI simply delusional, having drunk
the Itanic kool-aid? Or is it possible that SGI couldn't have built a
comparable system that could scale up to 256 CPU's (in the "extreme"
US$200,000+ configuration) using x86 technology? This is a company
that, despite its numerous business mistakes over the years, *does*
know how to build high-end massively SMP 64-bit UNIX systems.
>Porting VMS to a less than ideal architecture has already been done in the
>case of Itanic. What's the big deal in porting it to another less than ideal
>architecture (x86-64) if it will permit VMS to be more widely sold and used
>and ensure its survival? In the .edu space, free campus licenses of VMS on
"What's the big deal" is a rather casual remark to make; I'm sure it
cost millions of $$$ to do the Itanium port, so why should HP embark
on another porting project that would likely cost even more (given the
architectural weaknesses of x86-64 such as too few registers and
cumbersome 4-level page table mapping), and be even more complicated
for ISV's to port their Alpha apps to? This isn't something like
Linux where you can just recompile and then throw the code out there
and let your users be beta-testers -- this is a product that costs a
huge sum of money for HP to support, and each port to a new platform
represents a huge and ongoing expense in terms of long-term support
and maintenance commitments to their paying customers.
>The more people in business, government, and academia see the strengths of
>VMS, the more they will give their heads a shake and wonder what were they
>thinking before. It won't cause Microsoft to lose any sleep, but it will up
>VMS sales to the point where we, the VMS community, will not go to bed each
>night wondering whether our collective billions of dollars invested in VMS
>infrastructure and applications are at risk tomorrw from an EOL announcement
>of VMS from HP.
This may be true, but I don't think x86-64 is a feasible solution for
a commercial vendor and a "mission-critical" OS. I actually think it
was a mistake for Sun to port Solaris to x86-64 because most people
are just going to run Linux on their dual-Xeons and then steal
whatever code they want (like DTrace) from OpenSolaris and Sun won't
make any money from the deal.
-Jake
- Next message: David J Dachtera: "Re: Clusterwide QUEMGR file"
- Previous message: Wayne: "Letters to the new CEO"
- In reply to: John Smith: "Re: Could a PC do this?"
- Next in thread: Dave Froble: "Re: Could a PC do this?"
- Reply: Dave Froble: "Re: Could a PC do this?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
