Information Technology Reference
In-Depth Information
11.6 Conclusion
Cloud computing can benefit biology and medical studies by support-
ing collaboration with other scientists and providing cheap access to large
amounts of HPC resources. However, utilizing these HPC cloud resources
requires users to undertake a complex setup procedure that consists of cloud
resource selection, configuring cloud resources for HPC, and application
deployment. Researchers looking to take advantage of cloud computing to
carry out HPC analysis require an understanding of cloud architecture, soft-
ware to be run, and cluster management, which is beyond the scope of most
researchers. While e-science and research cloud solutions simplify access
and execution of applications on HPC resources, they do not solve the dif-
ficulties in developing and exposing analysis tools. As such, researchers still
require enough computing knowledge to utilize and manage large amounts
of HPC cloud computing resources, or they become reliant on financially
motivated cloud service providers. In response to this problem, the following
question was asked: How can a researcher with limited computing knowl-
edge become a cloud service provider?
An SaaS cloud framework was developed with the aim to simplify the pro-
cedures undertaken by service providers, in particular during service deploy-
ment and exposure. By identifying and automating common procedures, the
time and knowledge required to develop cloud services is minimized. Three
procedures were identified and became the focus of automation: applica-
tion deployment, interface generation, and service storage. By automating
application deployment, the computing knowledge required by biology and
medical researchers to access cloud software is reduced. By automatically
deriving an interface from the inputs and outputs of a service, the program-
ming requirements to expose software as a service are reduced. Finally, by
providing service storage, the time taken for analysis is reduced (through the
reuse of deployment information).
Implementation of the SaaS cloud framework was realized in the form of
Uncinus, a research cloud prototype compatible with Amazon EC2. Fulfilling
the requirements of automated application deployment required that cloud
resources be configured for HPC. To support this functionality, a new cloud
model called HPC as a service (HPCaaS) was proposed that automatically
configures cloud resources for HPC. To support automatic interface genera-
tion, an XML-based language was developed, as was a parser to translate
inputs and outputs to web interfaces. To fulfill service storage requirements,
an application broker was built for clouds that supported publication of cloud
resources and software services. By implementing features of the framework,
Uncinus benefits biological and medical researchers by simplifying the pro-
cess of developing and deploying software on cloud resources configured
for HPC. Cloud services can be built by publishing attributes (input/output,
computational requirements, etc.) through easy-to-use web interfaces.
Search WWH ::
Custom Search