Information Technology Reference
In-Depth Information
uniform. Hardware costs are reducing faster than
computer networking costs. This can be seen as a
threat to Cloud Computing, which is based on the
premise that buyers cannot afford for the hardware
costs but are willing to connect to the rented data-
center to have their jobs done and get the results
over the network. While the electronic costs are
reducing, there is increasing focus on electrical
and energy costs. In this day of price pressures
and almost free resources on the internet, (Durkee
2010) presents an interesting argument as to why
cloud computing can never be free. However, there
are ways to meet the price pressures at least partly
as the next paragraph indicates.
grammer can harness. Bugs cannot be easily tested
in local environments, so may have to be tested
and fixed in the cloud itself. There does not seem
to be enough support or tools for debugging and
development or even version control in the cloud.
Storage and representation of large amounts of
data is a significant problem. Currently, there are
only a few providers of the cloud infrastructure, so
discovery as a manual process is feasible. How-
ever, as the cloud providers grow in number and
services offered, discovery will become an issue.
As Richard Stallman of the free software
foundation fame feared (Michael et al. 2009),
there is a risk of the clients getting captivated
in the proprietary systems in the cloud, without
much recourse. This fear is not unfounded. Cus-
tomers participating in cloud computing can lose
in several ways (Durkee 2010), if they are not
careful. As of now, there do not seem to be many
attempts at standardizing the cloud environment
for portability. Not all applications yield to the
map-reduce framework that is typically used for
deploying applications in cloud. There is a need
to come-up with other techniques to deploy such
applications.
Solutions and Recommendations
We are optimizing the utilization of resources and
using otherwise idle resources in cloud computing.
So, there is not doubt that the overall cost benefit
analysis will favor cloud computing. Still, energy
costs cannot be ignored. Modern technologies
are amazingly cheaper than the traditional ones.
Following the trend is the data networks versus
electrical networks tradeoff. It is much cheaper to
transmit data (photons) on fiber optic cables than
it is to transmit power currents (electrons) on cop-
per cables. It is therefore cost-effective to locate
the computing machinery at places where electric
power is cheaper and utilize WANs to harness the
computing power remotely. This trend is already
in affect with quite a few companies locating their
data centers in energy efficient locations. This is a
clear direction that cloud computing should take
in future as well.
Solutions and Recommendations
Google's Bigtable (Chang 2010) is a popular
solution to the problem of data storage and repre-
sentation. The solution is designed to easily scale
to thousands of machines handling petabytes of
data. Parallel programming constructs must be
made available to the programmers to allow mas-
sively parallel computing in the cloud. There is a
scope of research in this area. Development and
debugging environments should be provisioned
in the cloud.
Standardization of the APIs and the cloud
environment is key to portability, mobility, and
wider usage of the cloud computing paradigm.
Eucalyptus (Nurmi et al. 2009) is an attempt
towards standardization. But there is a lot that
needs to be done in this area. Standardization
Deployment Issues and
Standardization
Cloud Computing is ideally suited for large ap-
plications requiring large scale parallel process-
ing. However, virtualization that is a necessary
ingredient of Cloud Computing limits the amount
of parallelism (Michael et al. 2009) that the pro-
Search WWH ::
Custom Search