Information Technology Reference
In-Depth Information
finance their work now sits in a room full of servers that are not always
used. Also, although some researchers are comfortable working with cluster
configuration, parallelization, and computer programming, that is not the
case for all of them. This lack of expertise means they have to pay a quali-
fied professional to handle all the initial configuration and maintenance of
their infrastructure.
6.2.6 Security and Reliability
One of the main concerns for these large-scale infrastructures is security.
It is a common practice to share computing resources among scientists inside
the same research group and with outsiders. Despite this shared environ-
ment, research work needs to remain confidential during its development,
sometimes because of applicable legislation, until the scientist decides it is
camera ready. Having this in mind, words like authorization, authentication,
confidentiality, and accountability appear right away.
Together with a secure environment, scientific executions require a highly
reliable platform. This is especially true when considering that some com-
putations can take weeks or even months to complete. Losing a month of
work just because of a server failure is simply not an option. This means
that a platform for scientific executions must have adequate mechanisms to
support these requirements.
6.3 Related Work
Approaches such as desktop grids and volunteer computing systems like
BOINC [4], OurGrid [5], Integrate [6], and UnaGrid [7] have laid the bases to
allow scientists to take advantage of large computing capabilities. Throughout
these kinds of solutions, researchers are able to access high-performance
platforms to run their workloads. However, the technical effort required to
run a defined workload under such conditions is generally too high for an
individual researcher with a tight schedule.
Recent developments in cloud computing solutions have aroused the inter-
est of the scientific community. Much effort has been expended to achieve
traditional cluster/grid performance in cloud environments. Some com-
parisons between cloud and grid have been made to show the benefits and
challenges presented by both technologies [8-10], in some cases combin-
ing them through a hybrid approach [11]. Results from important research
projects such as the Magellan report [12] have shown that cloud comput-
ing can fit scientific requirements under certain circumstances. Despite this,
the technical complexity of the configuration process is still high.
Projects like the NGS Portal [13] have strived to integrate domain experts'
knowledge into preconfigured application templates that are ready to run.
Search WWH ::
Custom Search