| |
Open Source Commentary from Navica's CEO,
Bernard Golden
April 2007
In This Issue
-
The New Economics of IT, Part II
-
Navica News
The New Economics of IT, Part II
Last month’s newsletter,
you’ll remember, focused on the factors wrenching the
old economics of IT: expensive hardware, complex (and expensive)
software, and organizational talent devoted to tweaking the
software infrastructure in order to get the last drop of performance
wrung out of it. All based on an assumption that the underlying
capabilities of the hardware and software components are inadequate
to meet IT’s functionality requirements; therefore,
endless tinkering is a fundamental requirement. In essence,
this assumption is that your IT infrastructure is like a Ferrari:
designed for maximum performance, staggeringly expensive,
and always in the shop for repairs.
As the newsletter noted, this assumption is all wrong for
today’s economics of IT. Unlike the past, today’s
hardware performance is incredible. In fact, we have too much
performance. Most data centers are stuffed with horsepower-laden
servers operating at less than 10% of capacity. Consequently,
operating a software infrastructure designed to wring every
ounce of performance out of puny hardware makes as little
sense as using a Ferrari to ferry groceries home from the
grocery story. Most of us make do with more different transportation
for this purpose: a practical sedan, a very practical truck,
or an ultra-practical minivan. The metaphor I used last month
was that we are treating our infrastructure like a thoroughbred,
when we should be thinking about it as a draft horse.
If you start thinking about your infrastructure like a draft
horse, or a minivan, you change your perspective completely.
Best-of-breed goes out the window; instead, you seek out best
value for money. You no longer assume a high performance infrastructure;
instead you seek out commodity hardware. And, crucially, you
avoid complex software solutions that require complicated
configuration and integration. You insist on straightforward,
self-contained infrastructure components that integrate via
standards-based mechanisms.
What is the order of battle for the new economics of IT?
First, commodity hardware. Many companies are already doing
that with respect to their servers. However, look to other
categories for their commodity equivalents, e.g., commodity
network switches, etc.. You should pay a category leader tax
only when you need the category leader. Otherwise,
you’re overpaying for what you need. Admit it, you’re
feeling queasy right about now, aren’t you? You’ve
delighted in jettisoning your Sparc servers in favor of x86
boxes. But, take away my Cisco in favor of 3com (or, to really
drive home the comparison, Linksys)? Take a hint from Google.
You’re not buying a durable, you’re buying a disposable.
If you ever outgrow your hardware, throw it away and start
again. If you need to dump your minivan for a Ferrari, go
ahead. But don’t buy a Ferrari today because you might
someday get a call to run at Monaco.
Second, inelegant software solutions. Admit it, now you’re
really queasy, right? Thinking about doing something inelegant
stresses every nerve in your body. I mean, we’ve grown
up with complex designs, five year roadmaps, elaborate architectures,
all done – of course! – in the name of busting
a gut to distribute pesky bits over a rickety infrastructure.
Well, if you don’t like the term inelegant, how about
a more acceptable euphemism: simple. Insist on standards-based
integration. Things like Web services. And XML. Both get criticized
for poor performance, as though poor performance means anything.
In a world of commodity, disposable hardware, it’s easier
to add more metal for performance than it is to optimize complex
software. And believe me, it’s easier to deal with a
sales rep from Ingram Micro than it is to deal with the one
guy in your shop who understands the hairball he put together
to string together two proprietary interfaces.
Third, issue different marching orders to your staff. Stop
taking elegant solutions from them. Insist on simple building
blocks that can have stuff replaced within the infrastructure,
not complex integrated stuff that requires a wholesale rip-and-replace
to change anything. Every time you let someone fall in love
with his (or her) creation, you’re signing up for a
lifetime re-enlistment. The approved term for this kind of
deliberate design is enterprise architecture. Make sure your
staff creates one and that it’s layered, uses clean,
standard interfaces, and incorporates only easily replaceable
components.
Fourth, just say no to proprietary sales reps. They dream
up reasons why you need to buy their latest and greatest.
And then you’re locked into their framework, their roadmap,
their need to meet quarterly estimates, their agenda. Just
look at Vista. Companies are having to develop three-year
strategic plans to deal with an OS upgrade – an upgrade
that delivers no discernable end user value, but does deliver
more ability for Microsoft to integrate Vista with a lot of
other Microsoft server products, that they’ll then try
and sell you on the basis of “works best with Vista,”
which “you’ve already invested in.” It’s
never-ending.
Fifth, open source. This doesn’t mean occasionally
considering it. And it definitely doesn’t mean evaluating
it by the standards of how you’ve done things with proprietary
software. I’m amazed at some of the comments I get to
my blog postings on CIO. People criticize open source because
it doesn’t “deliver business value.” What
they typically mean is that they’re used to letting
the vendor do their job of deciding what their infrastructure
should look like, then providing them a roadmap of their infrastructure
development plans, and then pre-integrating the solution with
the vendor’s favored software partners. So, naturally,
when you look at open source, it fails to do that. No open
source vendor is going to do a dog-and-pony show and then
build your proof-of-concept to get you committed to their
solution. Instead of asserting that open source doesn’t
deliver business value, run an experiment. Find out for yourself
what the costs of doing open source are. And besides, as open
source economics eats away at the margins of proprietary vendors,
they’re going to start looking more like open source
vendors. They’ll do less of the legwork for you. So
get used to thinking for yourself.
Sixth, reinvent your staff skill sets. If you’re going
to need to be more self-reliant, get ready. Bring more real
technical talent in the door. And, if you don’t want
to do that, hire some outside expertise. But be careful. Most
system integrators have grown fat and happy in the proprietary
world. They’ll be the first to tell you that open source
is a great vision, but the reality is that “you need
to rely on the established vendors.” They’re just
afraid for their business model. Unisys and CapGemini have
established open source practices. See what they have to say.
Seven, reinvest your savings in doing something new. Experiment.
You will save money with the above six items. Take the savings
and do some new things. This is critical. When the economic
realities of a situation change, it offers new opportunities
for those that grasp them. You have to go for it. Otherwise,
you’ll be stuck endlessly optimizing a complex, inferior
infrastructure. If computing is cheap, then it’s possible
to invent new business offerings that take advantage of cheap
computing. If you can’t find something better to do
with the millions you’ll save by saying no to Vista,
you deserve to be fired.
The old joke is there are three types of people: those who
make something happen, those who observe something has happened,
and those who say “what happened?” I’ll
add a fourth type, those who insist nothing can happen because
it’s too difficult to get started. In today’s
economy, this latter type is an endangered species with poor
life expectancy.
Navica News
You can hear me speak at these upcoming events:
May 10, 10:00 a.m.: "Open Source Virtualization: Creating
an Action Plan", Red Hat Summit, San Diego, CA
If you are interested in having me speak at your
organization:
Contact me directly via email.
|
|
