October 31, 2001

Egenera's Linux blade cluster: Great technology, but can they sell it?

Author: JT Smith

- by Jack Bryar -

Last week, I pointed out that many Linux and open platform developers
tend to be so focused on the technical aspects of the products they
sell that they ignore the requirements of their customer. Those customer
requirements may be couched in technical terms but the business issue
the customer is trying to solve is probably much more than a
technical issue. That's why "better" technologies frequently fail in the
marketplace. The customers aren't necessarily interested in a better
technology -- they want a better business solution. That's why, whenever I see a see a product that is technically
superior to any competing system out in the marketplace, I look at the
vendor's marketing plan before I get too enthusiastic.

This past week I saw one of those better products. And I looked at
the company's marketing plan.

The system in question is a Linux-based hardware platform developed
by a Marlboro, Massachusetts, hardware startup called Egenera Corp. Founded by Vern
Brownell, the former CIO of Goldman Sachs, along with his secretary
and his favorite IT consultant, Egenera has married Linux clustering
technology together with "blade" form factor servers and some
proprietary management code from Turbolinux to create the best
damn transaction server system I've seen in a long time.

It doesn't necessarily mean they are going to sell any of their
"BladeFrame" systems, however.

The company was founded by Brownell to solve a major technical issue.
In transaction environments, servers cannot be allowed to fail -- ever.
System availability needs to be 100%. There is almost nothing as
catastrophic as a system failure. Transaction environment managers
routinely double or triple the resources needed to service daily
transaction volumes in order to avoid downtime from equipment
failures or demand bursts. At the same time, this strategy for insuring
against system failure is bumping against another hard fact. The value of
each transaction is falling to through the floor. In market after market,
from retail grocery transactions to stock market trading-room floors,
the pressure to lower transaction costs is growing intense. Managers
need to find ways to drastically reduce processing costs.

What Egenera has done to solve this seems so obvious it's a wonder
everyone isn't doing it. The company has used Red Hat Linux to logically
cluster servers together to create a redundant, nearly crash-proof system with
mainframe performance characteristics. Impressive, from a cost a
performance perspective, but no major advance. Egenera has also used
tiny footprint "blade" servers, the better to stack together as many as 24
two- or four-processor blade servers at a time. This is also a good
idea, but not particularly innovative.

However, the company has gone another step and eliminated all the
connectivity hassles associated with bundling all these systems
together. What it's done is integrated all the connectivity into
the rack itself. The individual blades are little more than trays of
processors connected together in a "smart" rack bearing all the
controller and connectivity resources needed to make this rack into
a network -- in this case a built-in high-speed ATM backbone. There's no
fiddling around with connectivity equipment. Individual blades can be
hot swapped in and out in a couple of seconds. The whole unit can be
treated as if it were a single, inexpensive Linux mainframe made up
of pluggable processor trays.

The company has added to this virtual mainframe functionality by
bundling it with Turbolinux's PowerCockpit systems management
software. When Turbo released this Linux-but-closed-source product not long
ago, its marketing team talked about the software's ability to deploy
identical software configurations onto dozens of raw machines
simultaneously, and PowerCockpit could also perform administrative tasks on dozens of nodes simultaneously. That functionality syncs up well with the Egenera architecture, resulting in a system that can be treated as one logical unit, but which can be taken apart
and put back together while running .

So what is the problem? The problem is that the company wants to
sell these direct to end users, and as stand-alone products rather than as
part of an array of equipment that a vendor or an integrator might
configure into a "solution." That makes the company's prospects far
more problematic. Will potential customers understand how Egenera's
systems can solve their management issues? Why would a systems integrator
champion a system being sold direct rather than through the channel?

A potentially bigger problem for the company is getting its systems
in front of potential customers. The company has opted to keep its
sales force small in order to keep costs in line. Unfortunately, that also
means that the company's geographical reach is effectively limited,
at least for a while. While one could argue that there's plenty of
business for transaction systems just in metropolitan New York City,
strategically this is a real problem. Among the firms likely
competitors are companies like IBM, a firm with a global reach, direct and
channel sales programs, and a business-savvy solutions team.

So will better technology guarantee that Egenera succeeds in the
marketplace? We'll see.


  • Open Source
Click Here!