Information Technology Reference
In-Depth Information
hand it does not contribute to any of the hard problems of how to develop a
complex distributed application. Using PAAS, the cloud operator establishes a
new software layer with a dedicated middleware API (application programming
interface) and in this way abstracts away lower-level details. This facilitates
development of applications on top of the given platform, but on the other hand
it firmly restrics the types of applications to those supported by the platform.
Today, PAAS platforms are mostly targeted towards typical data-driven web
applications with an additional focus on support for non-programmer interfaces.
Summarizing the IAAS and PAAS characteristics, an important gap can be
identified for the systematical support for a wide range of enterprise applications
in the cloud. This gap is only partially filled by existing enterprise solutions like
application servers as these have not been conceived with cloud properties in
mind and do not allow transparently exploiting additional resources of the cloud.
To achieve the vision of a versatile private enterprise cloud two fundamental
challenges remain:
-
How to turn a highly dynamic environment consisting of volatile nodes into
a robust, manageable cloud infrastructure.
-
How to design and implement enterprise applications such that they are able
to exploit the cloud characteristics.
These two challenges can be broken down into a number of more concrete re-
quirements. With regard to the first challenge, the cloud middleware should
require minimal installation effort and zero administration effort for the single
nodes, otherwise it would not be feasible to include the many different types of
computers usually found in a company network. Furthermore, the operation of
the cloud infrastructure should not affect normal operation of the nodes, e.g. it
should not restrict the way, an employee uses her computer. Therefore, the in-
frastructure has to deal with dynamically appearing and disappearing nodes, as
employees turn on and switch off their computers. To support typical enterprise
applications, the cloud environment needs to support administration tasks also
for applications distributed in the cloud and therefore facilitate a transparent
management of distributed applications as a whole. Finally, the deployment of
applications should be ecient in terms of resource utilization, which requires
monitoring the available resources and reconfiguring the deployment structure
based on current application characteristics and infrastructure shape.
For addressing the second challenge, the infrastructure has to provide an in-
tuitive programming model for distributed systems, which facilitates building
applications such that they can be transparently partitioned and deployed in
the cloud infrastructure. The computing model should also support distribution
transparency in the sense of hiding complex distribution and concurrency issues.
In summary, this paper aims at developing a distributed computing infrastruc-
ture for private enterprise clouds that meets the following requirements:
1. Require minimal installation and administration effort for the infrastructure
2. Support independently operated nodes and dynamic environments
3. Provide an intuitive programming model for distributed applications
