OT: Cloud Computing is coming to ...

Ian Malone ibmalone at gmail.com
Tue Jul 20 21:38:59 UTC 2010

On 20 July 2010 18:21, Michael Semcheski <mhsemcheski at gmail.com> wrote:
> On Tue, Jul 20, 2010 at 12:51 PM, Marko Vojinovic <vvmarko at gmail.com> wrote:
>> Specifically, assuming that I have my own hardware to set the whole thing up,
>> what is the difference between having a server (possibly virtualized), and
>> having a server "on the cloud"? And what is the main benefit of the latter over
>> the former?
> Here's an example.  Imagine a University.  The School of Medicine has
> 10 servers, the Business School has 5 servers, the Engineering School
> has 15.  Additionally, Admissions has 1 server which is mostly idle
> for 9 months out of the year, but pretty busy during October, November
> and December.  The Chancellor has a server, HR has 5 servers, etc.  In
> total, there are 100 servers, but no one unit has more than 15.
> So, you end up with 20 different units each operating between one and
> 15 servers.  When they need to buy a server, they have to get the
> departmental IT guy to spec out a server and requisition it.  He buys
> one or two systems per year, and doesn't have a lot of clout or
> experience buying them.  He also has a bunch of other tasks - desktop
> support, application development, etc.  And he can't really know how
> much RAM, processor and disk they need, and its difficult to add more
> later, so they each overbuy a little.  Few of the departments have the
> resources to buy redundant servers, since in many cases they only need
> one.  Many of the servers end up sitting in inadequate space, plugged
> into the wall, without dedicated power management or cooling.
> With a private cloud, there'd be one group that does all the servers.
> That one group purchases the hardware autonomously, uses fancy tools
> to monitor the hardware, has a room with redundant power and redundant
> cooling.  Individual departments have a difficult time predicting what
> they need, but collectively its a lot easier to plan and make sure
> that everyones needs are met.  So there's less overbuying.  And the
> people managing the servers are only tasked with managing servers.

While there are quite a lot of advantages to this there are some
disadvantages, particularly in the research setting you describe where
each department will have its own set of grants which will have
allocations for computing resources and everyone will have a different
idea of the priority on their work. In reality departments that only
need a web presence usually use centralised resources anyway and
departments with heavy computing needs do have people with more
experience working on their planning. In a way this might be easier in
a business, where there's one pot of money which management decides
how to allocate. I'm sure someone with more experience in a business
environments will be along in a minute to say the opposite.


More information about the users mailing list