Putting your trust in another company to supply your computing
needs remotely over the
internet sounds daunting but managed well it can show
benefits.
Hardware as a Service is in its infancy but the core technology
has been developing since the early years of computing. Back in the
1960s and 1970s we had time sharing on mainframes and
minicomputers, which gave way to managed services. The late 1990s
saw the birth of the ill-fated
application service providers (ASPs) which failed to survive
when the internet bubble burst.
Now managed services and outsourcing are evolving into
cloud computing. The drivers of the market are improved
communication channels, a pay-per-access pricing model and the
efficiencies provided by technologies such as virtualisation and
federated storage. IT has become a utility that can be turned up or
down as usage fluctuates. Cloud computing also smoothes out the
spikes that can occur unexpectedly in IT budgets when a server
needs replacing or another problem pops up to drain resources.
Ted Chamberlin, research director at Gartner, says, "It's pretty
evident that we're still at the infancy of Hardware as a Service
but it is related to managed co-location, a very mature offering.
The majority of potential clients are server-huggers; they refuse
to lose their grip on their hardware. Co-location over the cloud
removes the worry about the biggest capital expenditure issue -
hardware procurement refresh."
Issues such as staffing, licensing, depreciation, maintenance
and back-up become the service provider's concern, leaving the
internal team to concentrate on the constancy of the data flow
assured by the service level agreement (SLA). IT managers will have
to find the best offer from established providers like IBM, HP, and
Fujitsu Siemens, competing with new entrants in the form of Amazon
and Google, and new services from AT&T, BT, Colt and other
telecom companies. Legal issues, such as compliance and the
physical location of data, will also have to be considered,
alongside which services can be trusted to the cloud.
Joe Duran, Primergy product manager at Fujitsu Siemens
Computers, explains, "The measurement of what was used and how well
that was provided will be key but there are different charge models
depending on the type of requirement you have. I don't think
there's going to be one absolute set of rules, just as there's no
single set of rules for how you buy your IT equipment."
Current suppliers of managed services are adapting to cloud
operations and working out what charge systems are suitable. Ian
Brooks, HP's head of internet and mobile strategy, says, "If it's a
storage service it's per Gbyte, or for computational services it
could be billing per CPU-hour or per node server if you have a
clustered application.
"Longer tenure could be based on service levels. For example, a
smaller company that can do without e-mail for half a day would
agree to a basic service level 1 but a bigger company, where e-mail
is critical, would pay a higher premium for a guaranteed service at
service level 3 or 4.
In the longer term, where you get very large multi-clustered
environments, you may well move onto a market dynamic where you buy
certain periods of time - similar to the futures markets for buying
airline seats or coffee beans."
The ability to build and tear down new systems without
interfering with internal hardware is making Hardware as a Service
attractive. A prime example is developing private clouds - internal
systems that use the cloud over the corporate network rather than
openly across the internet.
IBM has opened several cloud computing centres worldwide for
customers working out their cloud strategies.
Dennis Quan, IBM's CTO for high performance on demand solutions,
says, "In February, we announced our Design and Implementation for
Cloud Test Environments service because most of the work that goes
into getting a cloud service up and running is the final test and
pre-production phase.
"Resources can be made available quickly by allowing access to
those facilities running on our hardware. One of the benefits of
the cloud approach is that you only pay for what you use and the
service provider takes care of the details of the management of
those underlying physical assets and the growth required to sustain
the client base."
Apart from the trust required, a secondary consideration is the
speed. The maximum speed attained on the internet currently stands
at around 9Gbits/sec under ideal conditions. Naturally, this area
is the focus of much research, and an Australian team at the
University of Sydney has developed a chip that promises a
theoretical speed of over 600Gbits/sec but it won't be available
commercially for at least five years.