Information Technology Reference
In-Depth Information
of the cloud resources has a huge effect on runtime of an HPC application
(Church and Goscinski 2011). Non-HPC-enabled clouds are the ideal platform
for running embarrassingly parallel applications (e.g., common genetic
analysis applications such as mpiBLAST). Embarrassingly parallel applica-
tions can take full advantage of cloud scalability to reduce the time and cost of
analysis. However, performance studies (Goscinski, Brock, and Church 2011;
Expósito et al. 2013) have shown that, when running communication-bound
applications, clouds should make use of hypervisors with low overhead and
high-speed interconnection.
Once cloud resources have been selected, additional steps must be carried
out to enable the use of clouds as a distributed computing system. The steps
taken are dependent on the cloud service model. At the IaaS level, this involves
construction of a virtual cluster, compilation, and deployment of distributed
software. These tasks were previously the job of system administrators and
are beyond the scope of most discipline (even computing) researchers. PaaS
is aimed at developers, providing users with a development environment
and automating the deployment of resources. The problem of this approach
is that the user has limited access to development tools and programming
languages, thereby limiting the scientific applications that can be deployed.
At the SaaS level, the user is able to access HPC applications through graphi-
cal interfaces; however, the user is reliant on whichever cloud services have
been made available. In specialist research areas such as gene expression
profiling and drug discovery, such software would have expensive licenses
or not be readily available.
In summary, the complex process of selecting cloud resources and con-
figuring the cloud is beyond the scope of most noncomputing researchers.
However, a number of solutions exist to simplify the use of HPC applications
and cloud resources.
11.2.2 Solutions Supporting Research on the Cloud
An analysis of the current state of projects and development of computing-
based packages and tools to support researchers leads to two major areas:
(1) e-science tools based on web application programming interfaces (APIs)
and grids (clusters or clusters of clusters) and (2) research clouds, in particu-
lar research applications exposed as cloud services (SaaS). Through e-science
tools, researchers can run complex software without directly interacting with
computing resources; examples include HubZero, P-GRADE, and AGAVE.
HubZero (McLennan and Kennell 2010) is an open-source software plat-
form for creating dynamic websites that support scientific research and educa-
tional activities and promote scientific collaboration using primarily the grid
infrastructure. By using HubZero, a scientific gateway (website) containing
discipline-specific resources, including software applications as well as data
repositories, can be formed, and users of the scientific gateway can contribute
by putting their own applications and data into the gateway for sharing.
Search WWH ::
Custom Search