|
|
How do you know when it's time to own your own data
center?
Data Centers can range from smaller
rooms of a few thousand square feet to huge separate
buildings of several hundred thousand square feet.
When you are talking about the
smaller rooms, which I might define more as “server
rooms”, every company of reasonable size will own one.
The reason for this is simple – to manage a technology
infrastructure within a facility there are certain
pieces of technology that need a place to live. This
would include your network environment (core switches,
some access switches, WAN edge components security
components, etc.) There is also a minimum amount of
equipment that is needed to support the system side of
the network – Active Directory components (Domain
Controllers, DNS, DHCP, etc), file servers, and
localized support devices that produce too much traffic
to locate across a WAN (such as desktop image servers,
etc.). Unless the company is going to outsource
technology operations completely, this level of room is
typically company owned and operated.
Beyond the core infrastructure we now
get to the point of deciding the location of business
applications. If the amount of technology beyond the
core infrastructure is minimal (low number of in house
applications), I believe these will continue to reside
in the local data center with the core infrastructure.
As the number of applications
increases, a decision needs to be made as to where to
place them. I don’t see this as a decision between a
company-owned data center and cloud computing at all.
Companies will surely continue to own and host a wide
variety of local applications. There are several drivers
to where applications live. Specialized or proprietary
applications will often need to be kept in house, local
to both the IT support staff and the business users.
This can be due to network latency requirements,
transactional database requirements, the need to respond
to changes almost immediately, etc. Less critical
applications (from a network or business control
perspective) can be placed in outsourced or collocation
site.
I will assume your question is asking
about a larger data center that is not located with the
offices. The main choices are to build your own data
center or use collocation space. Managing the critical
infrastructure (power and cooling) requires a degree of
specialized skills that may not exist in a firm. The
company may not be interested in developing or acquiring
those skills. That would drive towards an outsourced
solution. Alternately, the company may need a degree of
localized control over the environment, feet on the
ground in the data center for IT management reasons.
This drives towards an in-house data center, but can
also be accomplished by collocating staff at the colo
site.
One of my clients is a large bank.
They separate technology into several groups,
proprietary and no-proprietary, technology whose
footprint varies over time and that which is location
independent. Their resulting strategy is that
non-proprietary technology that varies in footprint and
is not location dependent can be placed in non-company
owned data centers. The use of the outsourced
collocation site makes their data center environments
more agile. It does not necessarily make them cost less.
So, if I can take a stab at your original question; when
the local data center / server room can no longer
support expansion, and there is sufficient technology to
support an additional data center location it is time to
look for an expansion space. If there is a need for
extreme control over the environment, the skill sets are
available or can be added, the capital dollars are not
an issue, and a financial comparison between company
owned and outsourced locations is favorable, then the
company should consider building their own data center.
If a compelling argument cannot be made for an in-house
data center, the function should be outsourced and
reexamined on a periodic basis.
|
