Information Technology Reference
In-Depth Information
automation are becoming more and more popular. All these paradigms are
loosely related to each other and we can say there are lots of similarities
among them. To better understand the impact of system-level virtualiza-
tion on these new computing paradigms we explain utility computing
and on-demand computing briel y herein.
According to Foster and Tuecke [6], the term “utility computing” is often
used to denote both a separation between the service provider and con-
sumer and the ability to negotiate a desired quality of service from the
provider. Here the provider may be an organization's IT department or
an external utility provider and the coverage of the service may include
storage, computing, or an application. Similar services are provided by
Amazon's Elastic Cloud [7] that charges its user as per their use of CPU
power, storage, and so on. Similarly, Salesforce.com [8] provides a Customer
Relationship Management (CRM) application as an on-demand service.
On-demand is a term used to denote technologies and systems that allow
users or applications to acquire additional resources to meet changing
requirements in a broader sense than utility computing.
Investigating more in grid infrastructure, system virtualization is a level
below it and is designed to hide the idiosyncrasy of physical resources and
myriad of different software platforms. As we know, the responsibility of
virtualization is to provide isolation and provisioning. The task of grid infra-
structure is to manage resources above the virtualization layer. This task may
include and is not limited to provisioning (on higher level), monitoring, and
QoS. The layer above is responsible to provide application-level services based
on the grid infrastructure. These layers may include the workload managers
to efi ciently optimize the use of grids. This situation can be visualized in
Figure 16.1
,
which shows the architecture of current distributed environment.
Here, the top-level service is considered to be utility or on-demand as now
these terms/concepts are becoming more and more important.
The above model seems to be correct in the i rst glance. But in practice,
it has some uni nished issues. In an actual enterprise computing environ-
ment, different users may have different needs and may use different
computing paradigms such as service-oriented architecture, utility com-
puting, on-demand computing, cloud computing, and so on, each of which
has its own characteristics and structure while all use the same underlying
grid computing infrastructure to utilize distributed computing resources.
That is, the users should have the ability to manage their virtualization
environments, not a i xed mapping as given in Figure 16.1. This requires
system-level virtualization. Traditionally, grid computing is based on
service-level virtualization which is not well suited and not ready for this
shift. The gap between the current virtualization layer and the grid infra-
structure is big. This leaves application services to have an extra burden of
coni guring and managing low-level details when users require different
system coni gurations. To solve this problem, we should have another layer
of service that can well integrate in the current grid infrastructure and
Search WWH ::
Custom Search