Information Technology Reference
In-Depth Information
Cloud Data Center Construction
Contrary to their major differences in application and architecture, on-premises and cloud
data centers share the same planning and building considerations. You will need to plan for
the issues discussed in the previous sections for on-premises data centers plus issues unique
to cloud computing data centers because the latter are built and perform for a different
scale of workloads compared to on-premises data centers. The issues that should be consid-
ered for cloud computing are described in the following sections.
Location
How location may be an issue might not be immediately obvious. You might ask, “Couldn't
we just place the data center on our current building? We have some space.” The answer
would be no. As we mentioned, cloud data centers are created at an entirely different scale
than on-premises data centers. Traditional data centers are often built on 5,000 square feet
of floor space, but cloud data centers would require a lot more because they involve scalable
resources that are virtually infinite in the perspective of the user. However, the hardware
used to provide those resources is nowhere near infinite. That is why cloud service providers
like Amazon, Google, and Rackspace are constantly building new data centers, adding more
servers, and updating sometimes relatively new hardware.
Because cloud data centers have a global reach, somewhere in the future, cloud service
adoption will become so prolific that so much hardware will be needed to keep up with cloud
computing's promise of virtually infinite computing resources that a single 20,000-square-foot
data center would not be enough for the customers of a single service provider. This means
that hardware resources would have to be added regularly to keep up with customer demand.
In contrast, on-premises data centers are built for a certain capacity and are rarely upgraded in
capacity, just maintained and kept up-to-date.
Another consideration when choosing the location of the data center would be the tem-
perature of the environment. The cost of cooling data centers takes up a large portions of
the maintenance budget. So it would be in your best interest to select a location that has all
the required facilities and can also be used for ambient cooling, at least for part of the year.
The reason companies don't build their data centers in the Arctic and save almost 100 per-
cent in cooling is that there is no infrastructure to speak off, no power or networking infra-
structure. Plus, it's doubtful anyone would actually want to work in such locations unless
they are paid handsomely.
The idea here is to strike a perfect balance of utility infrastructure availabil-
ity, the likelihood of using alternative cooling methods, and an available
workforce.
Homogeneous Hardware and Simple Workloads
Unlike traditional data centers, cloud data centers cater to only a few dedicated applica-
tions that run on the servers. This means that the architecture would be uniform, and so
will software and the network that connects it all.
Search WWH ::
Custom Search