Virtualisation could transform the way companies deploy
and manage their computer resources but to succeed it will need a
standard operating system.
In the past five years companies have been buying
servers in an ad hoc manner, deploying them on a one server per
application basis.
Consequently, many companies have assembled large server farms,
which have an average hardware utilisation of 20% or worse. This is
a waste of money and a management headache.
The corporate infrastructure initiatives from IBM,
Hewlett-Packard, Sun Microsystems and Oracle are all quite similar
and aim to virtualise the hardware layer of corporate computer
infrastructure.
Virtualisation allows an application to access a pool of
hardware resources that are always on tap. In this way, users no
longer need to buy one server per application. Storage
virtualisation allows a user to access data in a standard way,
irrespective of whether data is stored on disc or tape.
However, there are other strong motivations for adopting
virtualisation, particularly the idea of being able to provide
computer resource as a service, dynamically. In other words, you
pay for what you use and you get what you need when you need it -
utility computing.
So companies, especially large corporates, are receptive to the
idea of having computer resources that are managed, efficient and
outsourced to some degree - which is what IBM, HP, Sun, and Oracle
are talking about when they refer to "grid", "utility" and
"on-demand" computing. Right now, what can be delivered does not
amount to wall-to-wall virtualisation, or anything like it.
So the question is: how is it ever going to be delivered, given
legacy systems, existing server farms and the enormous difficulty
involved in relocating applications in a heterogeneous network?
Blade technology, grid computing, automatic provisioning, storage
area networks and network attached storage will play a part in
this, but for it to work, and work well, it will require a standard
operating system.
This is a little like the idea of Java (write once, run
anywhere), but Java was about downloading applets. This is about
moving whole applications around. There is only one operating
system that could fill this role: Linux.
Linux is very popular and fast becoming the dominant server
operating system. Analyst firm IDC has predicted that this year the
number of new Linux servers will equal or possibly surpass the
number of new Windows 2000 servers.
Linux is also attracting large corporations. This year, Reuters
chose to put its Market Data System on Linux, Ford moved to Linux
with a series of server purchases, and there are reports that
Unilever is doing the same. These are strategic purchases from big
companies.
To this we can add the fact that many governments are showing an
interest in Linux. There are several reasons for this. One is that
open source is a guarantee against buried "espionage software".
Open source is also relatively cheap.
If you look around the world, China has its own version - Red
Flag Linux - and Malaysia is following in its path. Other Linux
proponents include Brazil, Japan, Peru, Korea, Pakistan, India,
Germany and Spain.
Another plus for Linux is that it runs in many different
contexts and on many platforms - from embedded systems and small
palm-top devices up to IBM's zSeries mainframe. It also runs on
just about every 32-bit or 64-bit chip and has become the
developer's platform of choice. It is the only operating system
that has such a broad penetration of the processor and server
market.
Because it does not belong to anyone, if Linux becomes the
standard operating system, then no specific supplier gets an
advantage from it and all suppliers can choose to embrace it
without worrying that the next release is going to poison their
business.
So in respect of virtualisation of computer hardware, the idea
is starting to emerge that you virtualise storage by using storage
area networks and network-attached storage, and you virtualise
server hardware by using Linux - making it feasible to switch
applications from one server to another automatically and
quickly.
Within this capability you can cater for failover and make
highly efficient use of resources. However, for most applications
you will need a standard operating system, because you cannot just
pick applications up and move them from one environment to another
without recompiling, unless the two environments are very
similar.
The need for a standard operating system becomes clear when you
consider why and how TCP/IP became the standard that enabled the
internet. It did not happen because it was the best option or
because it was purpose designed to run a worldwide network with
hundreds of millions of nodes, it was not.
It happened because it was the only reasonable choice at the
time. The same is now true of Linux as regards hardware
virtualisation. Irrespective of its other qualities, it is the only
candidate that fits the bill. If we are going to have a worldwide
network that behaves as a resource utility, then a standardisation
of this kind is both desirable and inevitable.
Linux does not solve all the problems of server virtualisation -
and there are many, including legacy hardware that will never run
Linux and legacy applications that will never run on Linux. But
this does not matter. They will get excluded from virtualisation
and will be superseded. Also, if Linux is thrust into this role
then the open source community will inevitably work on making it
even more fit for purpose.
As for other versions of Unix, and other operating systems, they
can probably accommodate a Linux personality, both for the purposes
of hardware virtualisation and because they will need to stay
relevant. Ironically, this capability has already been demonstrated
by SCO, which offers a Linux Kernel Personality within its version
of Unix, so it can run Linux applications. Both Solaris, from Sun,
and AIX, from IBM, have this capability too.
I believe Linux is set to become the standard operating system
and as it does it will radically change the way corporations deploy
computer infrastructure. Indeed, it will change whole market for
server hardware.
Robin Bloor is chief executive of Bloor
Research