Information Technology Reference
In-Depth Information
of software-as-a-service (SaaS) clouds include science cloud, a drug discov-
ery information management technology (Accelrys 2011). Thus, discipline
specialists now have access to on-demand, scalable, and pay-as-you-go HPC
facilities. However, while clouds alleviate the costs of procuring required
information technology (IT) resources, the cost and time of learning how to
prepare an HPC cloud and its applications remain a problem to many users.
The rest of this chapter is divided as follows: Section 11.2 investigates how
HPC applications are delivered on clouds. The types of HPC applications
that suit clouds are examined, as well as how these applications can be pro-
vided to researchers.
Section 11.3 presents a framework for publishing cloud resources and cloud
applications. This framework incorporates methodologies used by current
e-science and research clouds to simplify the development of SaaS applications.
Section 11.4 describes an implementation of the proposed framework. The
outcome of this section is Uncinus, a prototype research cloud that supports
the publication of software and cloud resources.
Section 11.5 consists of a case study in which Uncinus is used to analyze
genomic cancer data. By building a genetic profile of cancer tissues, the can-
cer subtype was identified, which has ramifications in providing personal-
ized treatment for cancer. Through this case study, it is shown how medical
software can be published, exposed, and deployed on cloud resources with-
out the need for complex deployment procedures.
Finally, in Section 11.6 a conclusion of the work carried out and an analysis
of achieved results are presented.
11.2 Delivering HPC on Clouds
To deliver HPC on the cloud, an understanding of the cloud, software to
be run, and cluster management is required. Using this knowledge, the
user must profile the HPC application to be run and select the right cloud
resources. This resource selection process has a large effect on the time and
cost of running HPC applications in the cloud. Once resources have been
selected, the cloud must be configured to enable HPC applications. Most
cloud resources that support HPC are provided at the IaaS level (in the
form of virtual machines). The user must be able to configure these virtual
machines into the form of a cluster, installing middleware and schedulers
and deploying HPC applications. The complexity of enabling HPC on the
cloud is beyond the scope of most biology and medical researchers. However,
solutions exist in the form of e-science applications and research clouds that
aim to lessen the computing knowledge required to carry out research on
HPC resources and the cloud.
Search WWH ::
Custom Search