Information Technology Reference
In-Depth Information
monitors the status of the resources known by GridWay. When it detects that there
are not enough resources to fulfil the demand, an automatic negotiation with previ-
ously registered Grid providers starts. The component only negotiates the final
number of CPUs and the price within the limits previously agreed and stored in the
server. These limits, which we call pre-SLAs, must be negotiated off-line by the
parties in advance and a contract must be signed. Currently, SLA Negotiation does
not provide electronic signature of the agreements.
Another important technical feature is that the workflows submit the jobs using
the Distributed Resource Management Application API or DRMAA (Rajic et al.
2008). It adds flexibility to the platform because it makes it independent of the final
scheduler. So, GridWay can be substituted by another scheduler which supports
such a standard. It facilitates the migration to another computing allocation and
submission paradigm such as Clouds or an installation inside a hospital in a cluster
with a usual local scheduler such as Grid Engine or Portable Batch System (PBS).
So, the platform can be easily adapted to the technical evolution and different work
environments, improving its adaptability.
9.3 Added Value for the User
The aim of the e-IMRT platform is to provide remote tools for radiotherapy based
on the use of vast computing power to improve the accuracy and quality of the solu-
tions. From this point of view, the provision of these services clearly benefits from
the usage of Grid technologies. Grid started in the data and computing domains
to permit the collaboration among partners, sharing their resources to tackle the
problem of storing and analyzing large amounts of data or to make complex simula-
tions. Now, the platform uses the developed technology to provide real added value
services to medical physicists to help cancer treatment planning. The usage of Grid
infrastructure opens this possibility because:
• It permits the execution of complex workflows that need a large CPU capacity
that is not usually available within the hospitals. By using Grid, medical physi-
cists can have access to this capacity remotely and, what is most important,
reduce the time-to-solution to a level that can be acceptable for their daily work.
The time for executing it with only 1 CPU makes unpractical to use this Monte
Carlo simulation technique for daily work. Only when more than 40 CPUs are
available, the execution time becomes acceptable for inclusion of this process in
the internal treatment planning protocols. This is a big infrastructure for a radio-
therapy department and it is out of scope of its business objectives. For example,
the execution time of a treatment verification can be reduced from 193 hours to
only 4 hours. This means, from more than one week to an acceptable time for a
day of work (see fig. 9.3). A further reduction of this time is still possible and
work is continuing to improve the scheduling of jobs.
Search WWH ::




Custom Search