Information Technology Reference
In-Depth Information
the features found in e-science packages: the application sharing supported
by HubZero, P-GRADE's low-level abstraction methodology, and AGAVE's
separation of scientific logic from computing. A potential cloud solution was
investigated through the development of a research cloud framework.
11.3 A Research Cloud Framework
Known issues can be resolved by developing a unified cloud framework that
allows researchers to easily deploy and expose HPC scientific applications
in public clouds as services (Wong and Goscinski 2013). Each of these SaaS
cloud services is such that it abstracts both the complex deployment effort
and the tedious command line execution style of HPC scientific applications
into just a web form for scientists to comfortably perform computational
research in the clouds.
The basis of our cloud solutions is a framework (see Figure 11.2) that aims
to deliver HPC applications to scientists as SaaS cloud services. This frame-
work enables two different processes: cloud software development and
cloud service publication. During cloud software development, each HPC
application is described by a set of attributes and their associated values.
Three major attributes of an application service are (1) a virtual machine
image where a targeted HPC application has been properly installed and
configured, (2) a web form where parameters for the HPC application
would be collected and then passed to the API of the HPC application, and
(3) a host location for service invocation, (e.g., SaaS resources). During cloud
service publication, these attributes are published to an HPC application
SaaS Cloud
HPC Resources
HPC Application
Service Registry
HPC Application Service
Publishing
Virtual
Machine Image
Ye s
Accessing
Web Form
Service
Discovery
User
No
HPC Application
Service and Web-
form Generation
Deploying
HPC Application
HPC Application
IaaS Cloud
FIGURE 11.2
An overview of HPC cloud framework.
Search WWH ::
Custom Search