Information Technology Reference
In-Depth Information
Fig. 9.2:
Schematic view of the treatment plan verification workflow
The second main addition to the platform was GridWay. The main aforementioned
services, optimization and verification, use remote computing capacity quite inten-
sively. For example, in figure 9.2 the workflow of the verification process is shown
(Gómez et al. 2007, Pena et al. 2009). It comprises several phases. Three of them
(1, 3, and 5) are executed locally because they need limited computing capacity.
However, steps 2 and 4 demand a large amount of CPU time but, fortunately, can be
split in many jobs and distributed to remote computing farms using Grid interfaces.
The amount of jobs and their duration depends on the type and definition of the
treatment plan and is calculated on-the-fly within the workflow. The optimization
process has a similar workflow, where part of the work can be done locally, but there
are compute intensive calculations that have to be done remotely to accomplish the
expectations of fast return of the end users. The Grid jobs are managed by GridWay,
a metascheduler that can submit the jobs to several Grids using different middle-
ware as Globus or gLite, and permits their usage simultaneously. GridWay has
been enhanced with a plugin (Broker GW-SLA) which uses the SLA Negotiation
BEinGRID component to add computing resources dynamically (Bugeiro et al.
2009) (for more details on this component see section 8.5 in chapter 8). This plugin
