Open Source Commentary from Navica's CEO, Bernard Golden
October 2006
In This Issue
-
Virtualization: The Cure for Open Source?
-
Virtualization Seminar: Limited Attendance
Opportunity for Virtualization Smackdown (NOTE: Microsoft
Added for Virtualization Panel)
-
Navica News
Virtualization: The Cure
for Open Source?
During the past few months, this newsletter has focused extensively
on the changes open source is forcing upon the software industry.
The August and September newsletters discussed the challenges
commercial open source companies face in their efforts to
build viable businesses upon free (as in free beer) software.
In summary, when organizations can access software at the
click of a mouse button without any need to spend money for
licenses and never speak to a sales representative, many of
the traditional enterprise software sales and marketing tools
lose their potency.
Of course, many of these open source companies believe that
the much lower cost of open source will enable vast expansion
of addressable markets; the lower price point of open source
will enable many more organizations to afford the product.
One often hears this characterized as “open source will
allow the SMB market to finally have access to enterprise-class
software.” So, according to this formulation, by selling
to so many more organizations, open source companies will
be able to build very significant revenue streams.
While this has a compelling ring to it – after all,
who doesn't want to see the little guy (in the form of the
beleaguered SMB) finally have a fair chance in life –
I must admit, I've often had a hard time convincing myself
of its reality. I just haven't been able to shake the feeling
that even though I'd like it to be true, it just doesn't hang
together in my mind.
The reason lies within the nature of open source itself.
Open source delivers great products with lots of functionality.
The fact that source code is available means that new functionality
can easily be added by anyone that concludes it's necessary.
Open source is typically very high quality; indeed, it's often
higher quality than its commercial counterparts, due to the
fact that so many more people inspect the code and exercise
the product.
However, most everyone recognizes that it falls short of
commercial software in the all-important “fit-and-finish”
category. By fit-and-finish, I mean ease of use factors. Intuitive
interfaces. Comprehensible documentation. Ease of installation.
Ease of configuration. In other words, those product aspects
that make it easy for an end user to get the product installed
and working.
In fact, some of open source's most positive characteristics
work against these aspects. Open source is very flexible,
open for integration, capable of addressing unusual end user
requirements. By contrast, commercial products often are difficult
to extend, offer limited ability to modify, and deliver only
specified functionality.
What this means for the end user organization is that it
must (typically) have a fairly sophisticated IT staff on hand,
or be prepared to pay retail rates for external service providers
to help it get the open source product installed and configured.
And this is where I have problems with the potential for
open source and SMBs (or, for that matter, other organizations
that have technical staffs limited in size or capability).
I described this conflict between the product need for sophisticated
technical users and the lack of technical capability in end
user organizations as an “impedance mismatch,”
which generally refers to the inability of one system to efficiently
take input from another, according to Wikipedia.
In this case, it refers to the mismatch between the unspoken
requirements for users of open source products and the reality
of end user IT organizations.
Open source is perfect for sophisticated end user organizations.
They have very talented IT staffs, often employ people with
intense intellectual curiosity who enjoy the challenge of
exploring new software products, and the willingness to invest
in technology solutions, seeing them as offering the potential
for competitive advantage.
On the other hand, SMB companies often fall short of these
capabilities. In fact, far short.
First and foremost, SMBs typically invest much less in IT
than their enterprise counterparts. Most SMBs are under intense
cost pressure, since they are less likely to be insulated
from competitive challenges by being a “platform company,”
or at least the leading entrant in a given industry.
Consequently, SMBs typically run their IT organizations as
lean as possible. While the pricing of open source fits very
well with SMB limited budgets, SMBs don't usually have IT
staffs with spare cycles to explore new solutions, especially
ones with associated learning curves. SMB IT staffs usually
come to work and find a whole series of conflagrations, keeping
them busy throughout the day fighting one fire after another.
Beyond the paucity of resources, one must also look at the
quality of resources as well. While large companies like Dow,
Goldman Sachs, and Genentech draw talented and ambitious staff
(including IT staff) to them with the promise of interesting,
challenging work, SMBs do not. One must be honest and recognize
that SMBs, generally speaking, employ, at best, the middle
of the bell curve, talent-wise.
In truth that may be generous.
I met with an SMB a couple of years ago to discuss the potential
for open source within the company. The IT director described
their current software environment, which was a melange of
small ISV vertical applications, Microsoft infrastructure
products like Exchange, and an assortment of string-and-sealing
wax home-grown applications written in whatever language the
employee or service provider was most comfortable with.
The kicker came when the IT director confided their desire
to move to a new version of one of their vertical applications.
“However,” he continued in a hushed tone, “that
will be quite a challenge for us. It uses something called
'SQL'.” This from an IT director! Talk about an impedance
mismatch!
This experience reinforced for me the fact that a significant
proportion of SMBs scarcely have the ability to keep a simple
infrastructure going, much less the time and ability to experiment
with open source.
Consequently, I'm sure you'll understand my lack of confidence
in open source being the perfect match for the SMB market.
On the other hand, SMBs have been so poorly served by the
software industry that open source represents at least the
potential for higher satisfaction. Vertical ISV applications
directed toward SMBs are often terrible: poorly architected,
obsolete technology, incomprehensible interfaces, backed up
with inadequate documentation, low quality support, and shambolic
sales representatives.
The key for open source adoption in the SMB market is that
it must be offered in a format that preserves its benefits,
while mitigating its shortcomings.
For open source to become an effective force in the SMB market,
it must:
• Be extremely cost-effective via no license fees
• Offer real flexibility in terms of use (ability to
install on multiple machines in response to changing business
conditions)
• Allow product modification to the minority of customers
that desire to extend via source code
• Enable integration through standards-based mechanisms
While mitigating these shortcomings of open source from the
SMB perspective
• The requirement that a capable, curious technical
staff be available to learn and experiment with the product
• Avoid the “free puppy” open source syndrome
of low initial cost, long-term high cost for IT staff or external
service providers
• The difficulty of installation and configuration associated
with products that assume a highly technical user base
Overall, it's clear that without a different delivery mechanism,
the much-heralded potential for open source in the SMB market
is likely to be constricted.
Which brings me to virtualization.
You can't go anywhere today in the IT world without tripping
over this word. It's gone from complete obscurity to over-exposed
celebrity in just 19 short months.
While most of the attention attention about virtualization
has focused on its potential for server consolidation and
its cousins, failover and redundancy, I believe that virtualization
offers an entirely new software delivery mechanism that will
– in the near future – transform the way end users
receive and implement software.
Virtualization offers software vendors and open source creators
the opportunity to deliver their product in a package that
significantly shortens the time needed to get it up and running.
Clearly, this is in the interest of the software provider.
First, it increases the potential for end users to enjoy the
benefits of the product more quickly, making them more likely
to become long-term users and, more important, satisfied customers.
Furthermore, virtualized packages will reduce the challenges
of installation and configuration, which make up a significant
proportion of all technical support questions.
But it's the end user that will really benefit from virtualized
delivery of software. Instead of installing packages (and
sometimes package after package), and then trying to learn
the configuration of the product well enough to get it up
and running, usually under time-constrained circumstances
that preclude really understanding what the configuration
is actually doing (not to mention often leaving gaping security
holes), a user will download a single image, plop it onto
an already-existing VMM (virtual machine manager), perhaps
fill in a few straightforward questions regarding IP address,
port numbers and so on, and then bring the new virtual machine
containing the application up.
It's easy to see that this will transform the job of the
end user technical staff. Crucially, it will dramatically
shift the skill set issue in favor of open source. Instead
of trying to learn new skills for every new product, it will
allow end users to leverage existing virtualization skills
repeatedly as new products are brought into the virtualization
infrastructure.
The use of virtualization offers the best opportunity to
to address the single biggest drawback to open source software
– its complexity. By delivering the product pre-installed
and pre-configured in a virtual machine, open source vendors
and creators will make adoption of their products much easier.
While virtualized versions of products will benefit all users,
they will be particularly valuable to the SMB market, since
they address the largest challenge SMBs face in open source
adoption.
The potential for this kind of software delivery mechanism
can be seen in the recent VMware Server appliance challenge.
VMware asked people to contribute virtual machines pre-configured
with applications, allowing them to be shared at no cost to
anyone who cared to download them. As an incentive, VMware
offered prizes for the most creative appliances contributed
during the contest.
The number and range of contributions was amazing. There
were many infrastructure-oriented machines donated, including
DNS (always a pain to get right), security (traditionally
difficult to configure correctly), and firewall (difficult
to configure and too important to get wrong) machines. That
doesn't exhaust the list by any means. One that caught my
eye was an Asterisk machine.
The VMware Server appliance challenge offers a glimpse into
the future of software delivery. People want to focus on getting
their jobs done, and installing and configuring software is,
at best, a necessary evil that is part of the job; at worst,
a horrendous time sink that delivers no discernible value.
Of course, it won't necessarily be simple to get to the future.
First and foremost, virtualization must become part of the
fabric of the computing infrastructure. It must go from being
yet another software product to be installed and configured
to something that comes as part of the operating system, ready
to do business right out of the box.
The second requirement is that end users learn how to use
and manage virtualization software. This may be a tough sale,
since the technology landscape is littered with products that
purported to simplify a difficult situation and ended up introducing
even more complexity into the mix. However, this should sort
itself out over time as organizational experience confirms
that virtualization is relatively easy to use (although there
may never be hope for our SQL-challenged IT Director!).
The third requirement is that open source product creators
and vendors will need to package and distribute their products
in a virtualization-ready form. This may seem straightforward
at first glance, but will require some engineering work: if
the product comes pre-installed in a default configuration,
it will probably be necessary to create an easy (i.e., graphical)
way to manipulate the configuration, since it will not be
a safe assumption that a user must have gone through a challenging
configuration process beforehand, thereby ensuring that he
or she is ready to tweak a configuration via a CLI or file
manipulation. Again, as virtualization becomes more accepted,
this kind of interface should become much more common.
Finally, of course, will be the challenge of how to get all
the virtual machines working together in a coordinated fashion.
It would certainly be less than desirable to have ten virtual
machines running on a physical host, each with its own copy
of Apache running, to take one example. Far better would be
a single copy of Apache running, either on the host platform
or within a special purpose virtual machine, to be used by
every other virtual machine. This may prove to be a challenge;
however, it is certainly not insurmountable, and is far preferable
to the alternative, which is a single copy of poorly installed
Apache being used by a multitude of applications on a single
physical machine.
It's early days for virtualization, but it seems extremely
promising. It holds the potential to solve or at least mitigate
the impedance mismatch between SMB requirements and open source
realities.
Virtualization Seminar: Limited
Attendance Opportunity for Virtualization Smackdown
I will be speaking at a half-day seminar on October 27 in
Silicon Valley as the featured speaker on virtualization.
The sponsor of the event, Teladata,
has very graciously offered to allow me to invite a limited
number of guests. As an added bonus, the day will close with
a panel on real-world virtualization experiences; VMware,
Microsoft, and Xen will participate, so there should be an
opportunity for a real dust-up. If you would like to learn
more about the impact of virtualization on data centers by
attending this event, please register here.
Navica News
You can hear me speak at these upcoming events:
October 4, 10:30 a.m. (PDT): "Open Source ROI",
OpenLogic Webinar -- View the recorded webinar here.
October 10: 1:00 p.m.: "Trends in Open Source: Evaluating
Open Source for Your Organization", Lecture to visiting
Norwegian ICT Delegation
October 11: 1:15 p.m.: "Open Source Maturity Models",
OSDL Face to Face Members' Meeting, Portland, OR
October 12: 10:00 a.m.: "Making Your Organization Open
Source-Ready", GOSCON Conference -- Register at the GOSCON
website.
October 25, 1:00 p.m.: "Using Virtualization in Your
Data Center", Data Center Decisions Conference, Chicago
October 27, 8:30 a.m.: "Virtualization: Opportunities
and Challenges", Teladata Data Center Education Seminar
-- register here.
November 7/8, 8:00 a.m.: Day 1 Chair and Speaker "Evaluating
the Business Models for Open Source Software in Mobile",
Open Source in Mobile Conference, Amsterdam -- Register at
the Informa
Telecoms and Media website.
If you are interested in having me speak at your
organization:
Contact me directly via email.
You might be interested in reading my blog posts
at CIO Magazine:
Grids: The Wrong
Answer to the Right Problem
Analyst Firm Gets
It Half-Right
GPL3: The Gloves
Come Off
|
