Information Technology Reference
In-Depth Information
into slots based on the processor cores per instance and number of cores
required per job.
The last solution is the H2D cloud (Brock and Goscinski 2012), which pro-
vides services to discover compute resources and deploy data and appli-
cations. This cloud platform is capable of utilizing both local and remote
computational services for single large, embarrassingly parallel applications.
In this solution, compute resources are published to a dynamic broker ser-
vice that monitors the state of available compute resources.
By providing scientific software at the SaaS level, it is possible to mini-
mize the computational knowledge requirements needed to access cloud
resources. SaaS eliminates the time-consuming tasks of software deploy-
ment and hardware setup/management, in particular resource selection and
allocation; however, due to the specialized nature of state-of-the-art research,
there is limited incentive for attracting cloud software service providers.
11.2.3 Conclusion
While originally their purpose was to support business applications, cloud
providers have moved to support HPC applications. These HPC clouds
have large amounts of memory, computing power, and high-speed network
interconnections. By using these clouds, users can access HPC resources on
demand without the need for supporting staff or purchasing expensive hard-
ware. However, to deliver HPC on the clouds, a complicated setup process
must be undertaken. Care must be taken to select cloud resources that suit
the HPC application being run. Cloud resources must then be configured
to allow HPC applications to be run; often, this involves the construction
and management of a virtual machine cluster. The computing knowledge
required to configure, access, and use cloud resources makes clouds unsuit-
able for the majority of researchers.
To support research on clouds, access to resources and complex software
must be provided to researchers with limited computing background. Two
areas that have shown success in bridging the knowledge gap between com-
puting and research are e-science and research clouds. The approach taken
by e-science applications and research clouds relies on the abstraction of
computing resources from the application logic. While the tools generated
from these approaches appeal to researchers, they are not an ideal solution
for specialized research. The development of e-science applications requires
a multidisciplinary skill set, while the research cloud approach relies on
financially motivated providers.
The investment required to develop services for specialized research areas
(with a limited market) is not attractive for service providers looking to make
a profit. Therefore, the solution is to devise a research cloud that enables
researchers to take the role of cloud developer. This research cloud should
implement scheduling and execution as well as enhanced features relating
to service composition and resource discovery. Such a cloud can incorporate
Search WWH ::
Custom Search