It's difficult to read anything these days on what's
happening in the world of ITC without tripping on a virtual this
and a virtual that. Indeed, Quocirca has researched and reported on
virtuality for some years now, and identified many of the problems
that users are now finding as they make their way through the
virtual minefield, writes Clive Longbottom, service director at
analystQuocirca.
One of the issues that comes up with virtuality is how to manage
the software that is needed within such virtual environments.
Initially, this was done through the use of "golden images", where
all the component parts of an application or service (operating
system, application server, the application itself, plus any device
drivers and so on) are built and stored ready for use as a single,
complete image. As time goes by, a library of these images grows,
and some form of library management is required. This library
manager not only has to be able to identify all the images that are
available, but also has to understand all the versions of all the
component parts, so that it knows which images need patching or
upgrading when a new patch comes through.
If you only have a few images, this is not a major problem.
However, many organisations suddenly find themselves with thousands
of images, many of which have the same operating system underlying
the image. In comes a patch, and several hundred images suddenly
need patching. Also, each of these images is carrying gigabytes of
the same information within it, which is wasteful in terms of
storage space.
Different approach
Quocirca has written before about how Microsoft is bringing a
different approach to the market with its Project Oslo procedural
engine, which will enable images to be built on the fly. However,
it still builds each image from complete stacks of components,
which in itself can be slow and wasteful.
A US-based company,
FastScale, is taking a
different approach. While still based on just-in-time construction
of images based on a procedural model, it takes an
Occam's
Razor view of how this should be done. For example, the vast
majority of operating systems have now grown up as bloatware,
carrying around heaps of excess data, information and services that
are not required for a specific task. For example, the "inf" folder
in my (somewhat battered) implementation of Windows Server 2003
contains over 150MB of "stuff" - some of which may be needed, some
of which may not. Similarly, the operating system starts up all
sorts of services on booting up - just in case.
FastScale takes a thin approach - it holds a database of
required components, and comes provided with a set of rules that
contain the contextual dependencies between most enterprise
applications and the operating system that will need to sit
underneath them. In this manner, it can ensure that only the bits
of the operating system and application that are required are put
into the image. Not only does this mean that you end up with an
image that can be put together far faster, but it is also highly
tuned, and so runs faster. Also, as it is a slimmer, more efficient
image, you can place more of these images onto an equivalent
platform than you could with full-fat golden images.
Shared benefits
FastScale also shares the benefits of any procedural approach to
building images on the fly - any patches or upgrades are applied to
the image held in the database - and even then, with FastScale
these are only "applied" virtually. Therefore, any applications
that are incompatible with the patch or upgrade can still load up
the original base image and run without any issues. This also
enables organisations to apply patches and upgrades when they have
done any retro testing - and the use of thin, efficient virtual
images can even speed this up.
We think FastScale's technology is pretty impressive, but it has
the likes of Microsoft, Oracle and
Procession hot
on its heels. It is only a small organisation, but already has some
large customers. However, FastScale could end up going the way of
another interesting virtual technology company -
PlateSpin, which was
acquired by Novell, and so lost a lot the independence that had
been associated with it.
Also, FastScale is currently only focused on server computing -
which is a pity. Its technology could revolutionise thin-client
computing, as it could improve image densities by three or four
times. But, this would require a concentration of effort from its
R&D group that could run the risk of defocusing the company
completely.
However, for those looking at large virtual ITC estates,
FastScale looks like it could simplify the management of images,
while also speeding up provisioning and improving image densities.
As far as Quocirca is concerned, it is a company well worth looking
at.