Segmenting Inside

The embedded Linux market is a tough one to segment, as show by how two of the top contenders try.


Once in a great while you see a company doing what would be sane in
other markets, but might be a Herculean improbability in their own.


Yes, this has to do with the Linux market.

Specifically this has to do with the embedded Linux market, a realm
so fragmented that ‘chaos’ is too polite a description. It is also one
of Linux’s silent success stories. Odds are that you are within five
feet of one or more devices that have embedded Linux inside. Glancing
about my office I count three (a printer, a router, and a cell phone,
though I suspect the hub and print server at Linux-based as well).

The embedded Linux market is fragmented along several vectors. The
primary vector of discord is the application. Router makers and printer
makers and cell phone makers have different interest and needs with
embedded Linux. A while back my neighbors at Wind River were toying
with the notion of creating an online community where users in the
different markets could share innovations in a non-competitive
environment, but that initiative seems to have fallen in the gutter.


Now MontaVista wants to do the opposite.

Ignoring for a moment the unfortunate aspect of having the word
‘vista’ in their corporate name, the folks at MontaVista have decided
that the proper approach to the market is to offer embedded Linux
packages tailored to different market segments. They are not tackling
the relative industries (routers, printers, cell phones, etc.). MontaVista is segmenting their embedded Linux offering by CPU/platform – Atom, PowerQUICC II Pro, PowerQUICC III, TI OMAP35x, etc.

Unlike the x86 server market, where use variations between box
vendors are relatively limited, the chip market for embedded Linux is
highly fractured. The differences are allegedly significant enough that
loading a Linux distro down with cross platform packages is a burden to
the buyers. MontaVista claims that many in the market buy an embedded
Linux package and then customize it to their platform before using in


Which seems very odd given that the use of Linux Inside is typically for the more primitive functions.

MontaVista is segmenting their product to match the chip-based segment of the market. Now segmentation is a Good ThingTM
for marketers to do. What I find curious is that the assembly of a
Linux package by CPU is a significant segmentation vector and that it
has taken this long for a vendor to segment accordingly.

Which means it may not be a prime vector for segmenting.


Over at Wind River, they segment based the category of final product in which their Linux will be embedded.
There are Wind River Linux distros for automotive devices, networking
gear, consumer products as well as several medical and military specs.
Instinctively this seems to be the more rational segmentation model.
Consumer devices need user interface packages (image a G-Phone without
the G-UI). Networking gear doesn’t need fancy UIs, but it does need
routing and network security functions that a consumer device might not.

The method to MontaVista’s madness may be in their new Integration
Platform (sigh, another use for the acronym IP). Akin to SuSE’s
openBuild system, the goal is to provide customers with ways of safely
and sanely customizing MontaVista’s core distro. This saves buyers the
pain of finding, including and removing parts of a Linux distro to make
it work for the intended application.

Here is a contrast in market approaches: Wind River has both a
general purpose distro and a string of special builds for different
industries. MontaVista has a general distro with some reconfiguration
for different CPUs and with a tool to tailor the distro to your
specific needs.


Which approach is better?

I’ll have to give the short-term nod to Wind River. Business in
competitive markets moves fast. Wind River provides products
pre-configured for various industries, and yet which can still be
tweaked by the customer (or by Wind River) if there is some exotic
need. This helps customer get their products to market faster and
possibly cheaper. If Wind River were to engineer an
openSuse/MontaVista-IP type system for customization, then they would
be hitting on all cylinders.

The marketing lesson herein is that segmentation is always driven by
the customer base, not the convenience of the vendor. Segmenting by
industry is a natural for many technology vendors, but it may not be
the viable for your products. There are two primary goals in
segmenting, which we’ll be happy to explain once we land you as a
client. Your segmentation model must meet these goals. If you don’t
then you will embed your company into the ground.