Cloud Computing and SOA Convergence in Your Enterprise: A Step ...
Cloud Computing and SOA Convergence in Your Enterprise: A Step ...
Cloud Computing and SOA Convergence in Your Enterprise: A Step ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Mov<strong>in</strong>g Onward 227<br />
Performance related to network latency typically arises around large data<br />
sets be<strong>in</strong>g sent to <strong>and</strong> from the cloud comput<strong>in</strong>g provider. The larger the data<br />
set, the more likely network performance issues will come <strong>in</strong>to play. You can<br />
take steps to mediate the issues around network performance by underst<strong>and</strong><strong>in</strong>g<br />
the network performance issues that may arise through performance model<strong>in</strong>g<br />
<strong>and</strong> test<strong>in</strong>g <strong>and</strong> by design<strong>in</strong>g the applications, services, data, <strong>and</strong> processes to<br />
m<strong>in</strong>imize the use of the network. S<strong>in</strong>ce you have to pay for b<strong>and</strong>width, you save<br />
money as well as make your overall architecture perform better.<br />
Network performance model<strong>in</strong>g is the concept of creat<strong>in</strong>g models that<br />
determ<strong>in</strong>e how the network will perform under an <strong>in</strong>creas<strong>in</strong>g load <strong>and</strong> us<strong>in</strong>g<br />
that model to determ<strong>in</strong>e saturation po<strong>in</strong>ts. It is a good idea to perform this<br />
step, at least at a rudimentary level, to underst<strong>and</strong> any network performance<br />
issues that will affect your cloud comput<strong>in</strong>g solution before they happen.<br />
However, given the “bursty” nature of the Internet <strong>and</strong> the wide variations <strong>in</strong><br />
network performance from time to time, this is not an exact science.<br />
As an option, some cloud comput<strong>in</strong>g providers allow you to leverage<br />
dedicated connections <strong>in</strong>to the data centers that host the cloud comput<strong>in</strong>g<br />
platform, provid<strong>in</strong>g a dedicated pipe between your enterprise <strong>and</strong> the cloud<br />
comput<strong>in</strong>g provider. This, of course, tremendously decreases the chances<br />
that you will have performance issues. However, it also <strong>in</strong>creases the cost.<br />
Platform performance centers on the performance of the cloud comput<strong>in</strong>g<br />
platform itself. Most cloud comput<strong>in</strong>g platforms leverage a shared, multitenant,<br />
virtual architecture. You have your own virtual space, or virtual<br />
mach<strong>in</strong>e, but you share processors <strong>and</strong> storage space with hundreds, perhaps<br />
thous<strong>and</strong>s, of other users on that cloud platform. It is conceivable that the<br />
cloud comput<strong>in</strong>g platform will become saturated from time to time, <strong>and</strong><br />
thus performance issues could arise.<br />
There is not much you can do about this, other than work with your<br />
cloud comput<strong>in</strong>g provider to <strong>in</strong>sure that you get the performance you require.<br />
Many enterprises <strong>in</strong>sist on service-level agreements, or SLAs. These<br />
are legal agreements that require the cloud comput<strong>in</strong>g platform to provide a<br />
specific level of performance.<br />
Other th<strong>in</strong>gs you can do to protect yourself <strong>in</strong>clude extensive test<strong>in</strong>g of<br />
the cloud comput<strong>in</strong>g provider at different times of the day before you accept<br />
the platform. Or, you can simply call exist<strong>in</strong>g customers of that cloud comput<strong>in</strong>g<br />
provider <strong>and</strong> ask them specifically about performance. There are no<br />
guarantees here, <strong>and</strong> one of the downsides of leverag<strong>in</strong>g a cloud comput<strong>in</strong>g<br />
platform is that you do not own or control those systems.