It Takes More Than a Fast CPU to Rule the Web
Written By: R. Krause
Published On: June 26 2000
It Takes More Than a Fast CPU to Rule the Web
NEW YORK and LONDON, May 16, 2000 - Compaq Computer Corporation today
announced an aggressive attack on the $32 billion high-end UNIX market
with the worldwide launch of its new AlphaServer GS series systems.
Targeting dot-coms as well as customers in telecommunications, high-performance
technical computing, healthcare and financial services, the new AlphaServers
are optimized to support the most demanding e-business applications and
customer environments. With over 200 systems already sold, Compaq anticipates
$1 billion in revenue from AlphaServer GS series sales during 2000
and expects to double its UNIX market share within target markets in the
next three to four years.
announcement is another significant step in Compaq's strategy to deliver
the best Internet infrastructure to our customers," said Michael Capellas,
Compaq President and Chief Executive Officer. "With more and more companies
facing IT performance bottlenecks and shutdowns, the importance of the
infrastructure has never been greater. Analysts estimate that 90 percent
of this infrastructure has yet to be purchased - and these demands are
causing us to rethink the way systems are architected, designed, developed
and deployed. The new AlphaServers, combined with Compaq Services
and solutions delivered with our partners, are building a new foundation
for the most demanding e-business environments."
First, a disclaimer: we have always been of two minds regarding the former
Digital Equipment Corporation (now part of Compaq) and its products. On
the one hand, they had excellent technology and high quality, robust products.
On the other hand, they "couldn't market their way out of a wet paper
bag". It is the latter problem that makes us a little peevish.
loyalists and especially those who need/want "Big Iron" have been waiting
a long time for Wildfire. Compaq talks about Wildfire becoming a major
player in Internet infrastructure, but it is not clear that they really
intend to go after Sun Microsystems (the current leader) in that regard.
Rather, the "focus" appears to be on the traditional strengths of the
former Digital Equipment Corporation (DEC) hardware - datacenter, high
performance technical computing (HPTC), and "internal" infrastructure
(e.g., e-mail, process management). This is not necessarily a bad thing,
just a defocusing of the "official" Internet infrastructure effort.
are concerned that Compaq's current Alpha marketing strategy appears to
be utilizing the same logic that failed DEC in the first place - to wit,
an overarching focus on "the need for speed". Although CPU speed can be
an important consideration (at least in the desktop space), and can help
with searching large databases, it is hardly paramount in Web infrastructure.
We would have expected Compaq to provide more Web-focused benchmarks such
as TPC or SPECweb, rather than SPEC CPU2000, as evidence of Wildfire's
value as a Web infrastructure component. It may turn out that Wildfire's
TPC-C or SPECweb99 power blows the doors off everyone else's server, but
how the heck are we going to find that out? Compaq should have had these
figures two months ago.
is anticipating $1 Billion in sales in 2000 for the product (including
storage and services).
the current pricing structure (priced from $100K to >$1 Million), we estimate
this means a sales volume of around 1500 systems for the year. This is
attainable, although probably a little optimistic: although IBM sold 720
of its RS/6000 S80 systems in the fourth quarter, Sun sold 255 of its
E10000 "Starfire" (Ref. TEC News Analysis IBM's
Unix Servers Eclipse Sun). This means Wildfire needs to sell about
twice as fast as the Starfire - not impossible, but not the kind of odds
we'd take, even with the former-DEC installed base (although we hope they
the GS 80/160/320 do not meet their sales numbers, we expect a one or
more VPs to lose their jobs. It's not always fatal to be late to market
(although it is very painful); it's not always fatal to blow the sales
projections; but in this case, we believe it will fatal to do both.
These systems are not for Mom-and-Pop shops or small ISPs. With a base
sticker price of $100K, these systems are for larger companies, especially
those who expect to expand (the idea being one can expand from eight to
32 CPUs relatively easily). Alpha CPU history indicates that the machine
will perform well, although the per-transaction cost (i.e., in a transaction
processing situation, or a large database scenario) will likely be high.
For example, when running the TPC-H benchmark test (for "decision support
environments where users don't know which queries will be executed against
a database system" [Source: TPCC]), the IBM RS/6000 exhibited a greater
than 3:1 price/performance advantage over the previous large AlphaServer.
The AlphaServer had been tested six months prior to the RS/6000, so it
is likely that the Alpha's price/performance figure improved, but three-to-one
is a lot of gap to cover.
those companies with enough money to handle the expense, history indicates
that the Alpha-based systems produced by the DEC arm of Compaq will be
reliable and robust. Customers often use them for 24x7 operation. Although
we expect this trend to continue, potential customers should also exercise
the normal amount of caution that would accompany the purchase of any
"first edition". Customers should also ensure that their desired application
has a "Tru64 Unix" version (Compaq's variation of Unix).