Information Technology Reference
In-Depth Information
services registry. The HPC application services are proposed to be stored in
such a manner that their discovery and selection are easy. This implies that
the invocation advice and at least two attributes of an application service,
its host location (SaaS cloud) and its web form, must be published. It is pro-
posed to employ a dynamic broker (Brock and Goscinski 2009; Goscinski
and Brock 2010) of resources and services to publish application services,
thus allowing other end users to learn of the newly deployed application.
11.3.1 Framework Scope
In response to the major problems faced by discipline specialists in using
HPC clouds (see Section 11.2), three services are proposed: (1) automation of
HPC application deployment, (2) automation of HPC application service and
web form generation, and (3) HPC application service registry and formation
and application of a wiki-like knowledge base for interface regeneration and
HPC application customization. These research areas and the relationships
among them form our cloud framework.
11.3.1.1 Automation of HPC Application Deployment
To enable SaaS development by biology and medical researchers, there is a
need to automate aspects of the HPC application deployment. Supporting
this application deployment process requires at least two levels of abstrac-
tion: (1) low-level deployment that consists of methods to install and config-
ure an HPC application in a virtual machine and (2) high-level deployment
that consists of methods to save an image of the preconfigured virtual
machine and construct an API of the HPC application that together form a
deployable unit.
Low-level deployment focuses on automating the installation and configu-
ration of any software application on any computer system. The most com-
mon approach taken to automate low-level deployment is seen in package
management systems such as the Advanced Packaging Tool (APT) for the
Debian GNU/Linux distribution and its variants (Calinou 2012). These tools
automate the retrieval, configuration, and installation of software packages,
either from binary files or by compiling source code. However, such methods
are focused on single machines and are not designed for remote installation on
HPC clusters. To enable automated deployment of HPC applications on clouds
requires secure access to remote resources and automated resource selection.
High-level deployment focuses on automating virtual machine construc-
tion and configuration. Methods encapsulate an HPC application into a
virtual machine image and an API of the HPC application; this will form
a deployable unit that can be exposed and easily accessed by users as an
SaaS cloud service. To support high-level deployment in the cloud, there is a
need for tools that can set up an HPC execution environment consisting of
Search WWH ::




Custom Search