25
OCTOBER,
2004
"Meeting
SLAs is now the important goal"
Thus
spoke Nick van der Zweep when making one specific point but, indirectly
making another of significant importance in Service orientation. As director
of virtualisation and utility computing at HP, van der Zweep has an obvious
vested interest, namely that virtualisation technologies and systems now
play a crucial role in building infrastructures that are actually capable
of meeting SLA requirements. This is certainly true, especially as such
infrastructures get bigger, but it is now only a part of what constitutes
the process of meeting the needs of an SLA, as conversations between Solutions
Architect and Thomas Meyer, of IDC, show.
Virtualisation – basically the ability to 'build' and then 'remove'
a number of virtual server systems out of an array of available physical
servers so that changing resource requirements can be easily and flexibly
met – is fast catching on and is probably the key physical component
in meeting the of an SLA. If the service requires 100 servers for one day
a month, and only 2 servers for the rest then buying 98 largely unused servers
is wasteful and, these days stupid. Instead, virtualisation allows servers
to be redirected to high-priority tasks as and when required, which means
that the processor demands implicit in any SLA can be met in a flexible
and cost-effective fashion.
With Intel-based server architectures now becoming the norm, virtualisation
is also becoming easier to engineer. HP, for example, has formed a deal
with VMware, probably the best known supplier of virtual machine technology
for Intel-based systems. In this HP is actually following IBM, which struck
such a deal with VMware over a year ago. But the company is not standing
still on this subject. "Now we are working with Intel on building-in
virtualisation on-chip for the next generation of processors," said
van der Zweep.
This has considerable potential cost benefits. "Not only can users
save on the number of actual servers they require, particularly as the number
of processor cores per chip goes up and the number of processors per server
goes up," he said, "but they also save on the total cost of licences
required." There is an interesting side issue to this, which he acknowledged
as one to be faced by the software vendors, that their potential future
revenue streams could
But virtualisation, though important, is only one aspect of producing a
service. Indeed, as conversations with Thomas Meyer of analyst company,
IDC, show, the whole structure the concepts underpinning IT in its broadest
sense are in the process of inverting. The vendors have looked at the world
(and arguably many still do) from a purely technology perspective. They
have technology capable of producing some specified results, and it is up
to users to fit their business models to match. So the view was from the
technology out towards the users. That is now inverting, and it is what
the users want – the services themselves and the service levels –
which are becoming the key drivers. What is more, it is now becoming possible
to define a number of points in the service `stack’ to which users
can identify the modules that go make the services they require, and to
which vendors can work towards building the standards that will let them
interoperate.
So, though Meyer agrees the list is as yet incomplete, the elements of the
service stack start to shape up. Service level management is now at the
top, the core capability in meeting the needs of an SLA. Next down the stack
comes business process management, closely followed by business process
architecting – building the services from applications and utilities
(both new and re-used) and managing them in ignition, operation and tear
down. Next up is delivery management, which is what most now call web services,
the tools that deliver services to end users and information from them when
necessary. Underneath this comes the broad spread of infrastructure management,
which will include not only the management of the servers, storage and networks,
but also the applications and how they surface, plus their accounting and/or
licence management systems. At the bottom (but no less important for all
that) comes interconnect management – the actual physical servers
and the like on which it all runs. These must interconnect in a fashion
that will give the phrase `plug and play’ some real meaning. This
must become an anything-to-anything `no-brainer’ in a service-based
world.
Some of these will become extremely significant and some may not stand the
test in time but, as Meyer agrees, a top-down stack along these lines is
now the way users and vendors should be approaching all new developments.
www.hp.com
www.vmware.com
|
|
