Information Technology Reference
In-Depth Information
terminate resources when jobs are complete. This HPC software library
interface is made available through the application broker.
On the top (SaaS) layer, a service for supporting HPC application deploy-
ment was developed as follows: First, an API of the HPC application was
constructed. This API acts as a program stub for its corresponding HPC
application, which, when deployed, is installed in a virtual machine on the
Amazon EC2 cloud. Second, a web form exposing the HPC application ser-
vice is generated. This HPC application service can then be published to the
applications broker to be exposed as a software service. It should be noted
that each HPC application service would access the HPC services in the
HPCaaS layer and the HPC application installed and stored in a VM image
at the bottom IaaS layer through its web form.
The components that make up the framework are described in more detail
in the following sections. Section 11.4.2 presents the operation of Amazon EC2,
which provides cloud infrastructure services to the research cloud prototype.
Section 11.4.3 describes the construction of the HPCaaS model and how it is
able to abstract Amazon EC2 resources. Section 11.4.4 describes the construc-
tion of a research cloud called Uncinus; this cloud provides an application
broker to deploy and expose applications. Uncinus also exposes the services
provided by the IaaS and HPCaaS layers through easy-to-use web interfaces.
11.4.2 Amazon EC2: The Public IaaS Cloud Service Provider
Amazon EC2 provides various computer instance types specifically designed
for running HPC applications. Our work has utilized the elastic compute
cloud services and the elastic block store services to deploy and run HPC
applications. The simplest way to use the EC2 services is by accessing the
Management Console (Amazon Web Services [AWS] 2013). After logging on
to the AWS Management Console, a user can carry out HPC activities with
EC2 by performing the tasks of
1. Selecting the desired Amazon Machine Image (AMI) and launching
computer instances;
2. Installing and configuring software;
3. Establishing connection to the computer instances, running applica-
tions, and handling data transfer;
4. Terminating computer instances and releasing resources.
This approach of accessing a public HPC cloud service is quite ad hoc
and could be tedious for discipline scientists who have little background in
HPC. On top of the work in launching, connecting, and terminating AWS
computer instances, discipline scientists are also forced to deal with many
details to set up and configure an HPC cluster and install middleware and
software applications before the system is available for any actual scientific
Search WWH ::
Custom Search