Firms Cut Their Teeth on Blade Servers

In the high risk world of high-tech, businesses and the engineers who
develop new infrastructure technologies for them are forever faced with the
challenges of making smaller products that conserve energy and space, while
trimming costs. In enterprise computing, that is where blade servers
come into play. As with many technology issues, the
specter of debate hovers over all.

A blade server is an entire server that fits
on a single card, or blade, which means that network interfaces, the CPU,
the memory, and the hard disk are installed on the card. The blades are
plugged into a single chassis, where an IT manager can generally fit 16
server blades into the space previously occupied by a single server.

As opposed to commonplace mainframes and other refrigerator-sized server
systems, but very much in line with what major high-tech firms are leaning
toward, blades use low power processors to reduce power consumption and
heat, both of which save money and wear and tear on the system. Blades also
cut out the messy Ethernet cable aspect of the serving chore. The smaller
form factor is a plus, but density, failover (if one
server goes out, can another one kick in seamlessly) and price performance
are, for the most part, what blades strive to improve upon.

Blade servers are popular among ISPs and ASPs, and for applications such as
e-mail, Web hosting and domain name serving. But as the technology improves,
meaning as hardware and software for them evolves, you’ll probably see a lot
more in finance fields for banks, or in life sciences fields, where genome
studies require large amounts of clustering. With a market potential of
billions of dollars (IDC estimates $3.7 billion by 2006), one thing blade
servers are not, is hooey.

In the beginning…

Now, while you might have seen a number of announcements from Compaq , HP , IBM
, Dell , and Sun Microsystems about their new blade server systems, the tale of the market
niche goes back a few years, before the bottom fell out of
the dot-com market and high-tech firms dependent on hardware and software
infrastructure to help them run and expand their businesses started cutting
back on expenses and becoming leaner.

In 1999 and 2000, a slew of startups launched, promising that the bold new
form factor would prove to be more than just a novelty. Now, months away
from 2003, you’d be hard pressed to find any of these beyond Marlborough,
Mass.’s Egenera and Houston’s RLX Technologies, both of which seem to be
succeeding. Egenera’s high-end approach to blading (Credit Suisse First
Boston replaced 20 traditional RISC-based servers with Egenera’s BladeFrame,
which it uses to process more than 60 million financial transactions a day)
has been praised by analysts, while RLX has found additional funding in a
time when funding for IT infrastructure is almost nil.

Analysts such as Summit Strategies’ John Madden and IDC’s John Humphreys,
said a lot of the startups both native and abroad went under because of bad
business models, or because the “partnership ecosystem” was not ripe.

But something in the way RLX and Egenera moved caught the attention of the
major players, including Compaq, HP, IBM, Dell and Sun. They have all
announced intentions to get into the market. And why not? Research firm
Gartner Dataquest said last February that worldwide blade server shipments
could grow from 84,810 units in 2002 to more than 1 million by 2006.

RLX and Egenera, who it should be noted approach blades differently,
(Egenera shoots for high-end servers and envisions them powering the largest
of enterprises; RLX goes for datacenters at the lower end) are ramping up
product cycles and are in their fourth generations of blades, while the
titans are in first or second phases.

Blade server players

Among the big boys, HP tapped the keg at the server blade party in December
2001, when it unveiled its new blade server product series and began taking orders for the server
once code-named “Powerbar.” IDC analyst John Humphreys said at the time that
the “blades market is poised for rapid growth and HP’s approach to blades is
leading the way for the expansion of that market.” He also noted that HP had
cultivated well-developed ecosystems of partners.

Compaq followed
suit in January 2002. Launched in New York City, the ProLiant BL e-Class
server blade, code-named QuickBlade, promised businesses that they may pack
up to 280 servers into a standard 42U rack with hot-swappable fans and power
sources. Up to 20 blades can fit in a 3U rack, and those 20 blades only
require 7 cables as opposed to the 80 now required. Later, it launched the
BL p-class, with more power. While the e-Class served static Web pages, the
p-Class promises to enable dynamic Web hosting, terminal server farm and
streaming media.

Now, you would do well to ask whether or not HP and Compaq’s server blade
lines overlap, as Compaq remains a separate brand under the HP umbrella.
Could one line cannibalize the other? A fair question, and one that
ultimately remains to be seen. Just note that HP has designs to sell to
telco and network service provider customers; Compaq aims to power front-end
to high-performance SMP servers for multi-tiered environments. HP recently
refreshed its server blade family with the ProLiant BL p-Class system,
its first dual-processor blade server to hit the market. This is big news it
itself, as server blades are expected to scale out beyond traditional 1
processor to 2 and 4-way systems (or more) for added power.

After those two, you can count Dell and IBM as the next entrants in April
2002. Dell announced its new PowerEdge 1655MC, which holds up to six servers with two Intel
Pentium III processors in one rack, while IBM introduced its eServer BladeCenter, systems powered by Intel Corp.’s DP and Itanium
chips in addition to IBM’s own POWER server semiconductors.

Sun Microsystems is the remaining holdout, with products planned to go live
in the next few months. Sun plans to release two types of blades later this
year, including one that uses Intel chips and the Linux operating system,
and one that employs its own UltraSparc chips and Solaris operating system.

Ashley Eikenberry, group product marketing manager for Blade Computing at
Sun, told internetnews.com that Sun sees
the next wave of blades to be “media blades” to facilitate business
transactions.

Fine young cannibals
Egenera and RLX have already been touched upon. Both are understandably
proud to be surviving, early movers. Both are confident that they are far ahead
of the competition. Egenera just last week launched its new two-way
blades powered by new Intel 2.80 Ghz chips, while RLX unveiled its fourth-generation ServerBlade 1200i, in addition to a new control tower and clustering software. Egenera’s value proposition
is different from the other players. For one, it’s currently positioned to
run Linux, with support for Microsoft’s Windows .Net server coming in 2003.
For another, Susan Davis, Vice President of Product Marketing & Management at
Egenera, noted the other firms aren’t capitalizing on the potential that
blades have to offer.

“A lot of competitors want to offer very dense blades with a lot less
power,” Davis told internetnews.com. “That was never our focus. Our
goal was: ‘How do I take datacenter infrastructure and move it to cheaper
environment without giving up reliability and manageability.'”

Davis said the chief reason why major companies are concentrating on the
low-end is that they want to keep blades away from their high-margin
servers. Again, cannibalization is the driving fear.

Although RLX, like the major vendors, focuses on low-end, Bob Stearns, a
former chief technology officer at Compaq, now a board member of and
investor in RLX, said the larger companies should be fearful. He believes
blades will ultimately replace large servers.

“I see server blades replacing servers in many instances if the applications
scale out,” Stearns said. “It’s possible for high density concentrations of
servers to replace older types of systems. And any company that doesn’t
cannibalize themselves gets cannibalized by competitors.”

Moreover, Stearns is happy that the major players have come to the blade
server table, claiming that it validates what startups like RLX have been
working on for years.

Analysts don’t seem to believe firms will be any worse for wear for being
late entrants. After all, there is still the issue of standards to be hashed
out.

One standard to rule them all; one dissenter to question them

Last February, Gartner Dataquest published a host of blade server studies.
One of the issues that stood out is that of standards.

“A lack of standards will be a primary market inhibitor as many end users
will be reluctant to install a blade server that appears to be proprietary.
This restriction on blade server demand will encourage the development of a
standard designed specifically for blade servers to which the worldwide
server vendors adhere,” said Jeffrey Hewitt, principal analyst covering
servers for Gartner Dataquest’s Computing Platform Worldwide group. “The
acceptance of such a standard should help reduce end-user inhibition to
install blade servers.”

Currently, the standby is the CompactPCI standard.
Developed in the mid-1990s, this once-solid standard is widely acknowledged
by analysts and industry players to be almost Jurassic when folks talk about
what they expect to get out of blades. InfiniBand is another standard garnering much attention as a
possible replacement to PCI delivery, but this has yet to gain the momentum
experts expected of it.

And while many analysts are bullish about blade servers’ prospects, others,
such as John Enck, VP & Research Director of Server and Directory Strategies
at Gartner, are careful in what they tell their clients who inquire about
the hype.

“The market for blade servers is very immature right now,” Enck told
internetnews.com. “Right now customers are looking at what’s out
there and it is very confusing, with HP, Compaq, Dell and the rest of them all coming at this
from radically different angles.”

How good are the blades that are out there? Enck said as far as products for
plain ole Web serving, that most are adequate, but as far as handling a
complex workload, most are not up to it.

“Much of this is because the software is not equipped to do the provisioning
that is necessary,” Enck said. “Now, these companies have the right idea, it
is just very difficult to navigate this. I tell my clients the 2-year ROI
[timetable] of this is worth waiting on.”

Enck said there are also misconceptions. Without picking on any vendors
technology in particular, he said the idea that blade servers are much less
power consumptive isn’t true.

“Do the math,” Enck said. “When you load up a rack server with blades, the
power output is actually higher [than large servers] because the density is
greater. Right now, it’s just not panning out. Everyone is trying to find a
general value proposition.”

Enck also took exception to the fact that many vendors are downplaying the
importance of standards for blades, which he said can be attributed to the
fact that, of course, the vendors love their own proprietary technologies.

“Standards have to make it useful,” Enck said. “It’s not just about being
able to put HP blades in an IBM chassis. Look at the Intel-based server
market. It had no momentum until vendors started to agree. With blades,
vendords need to find the right connection for blades. Different designs
sport different standards and if they can’t agree what interconnects them
there will be no maturity.”