Information Technology Reference
In-Depth Information
final cost and price of such treatments. For those that do not want to use the pay-
per-use model or cannot access the Internet, it is possible to install the software (and
hardware) internally, although losing the flexibility of the external provision. Viable
because although there are strong competitors, there is room for new ones and, even
having a modest market penetration, the business could be profitable. In fact, one
small company can generate benefits executing only few thousands of simulations
per year. This is a small fraction of the total cancer cases, which are measured in
several million per year. However, the new company must surpass some important
barriers, as explained before.
9.5 Outlook
We strongly believe that the future of radiotherapy treatment planning must be
based on open solutions which will be provided via the Internet and that will
require high computing capacity. Both services and computing could be provisioned
on-demand with quality of service and improved security. The proposed model has
shown that these services demand a high number of CPUs for returning results in a
reasonable time. According the Directory of Radiotherapy Centres (DIRAC) of the
International Atomic Energy Agency (IAEA) (IAEA 2009), there are 6214 radio-
therapy centres in the world, with 7168 Linacs and more than 6000 TPS. Just in the
European Union, there are 980 radiotherapy institutions with a mean of 1.9 Linacs
per centre. The country with more Linacs is United States of America (USA) with
more than 2100 (2114 in 1870 institutions) followed by People's Republic of China
(981) and Japan (842). The ideal objective is having 4 linacs per million inhabitants,
so we can expect that the number of centres will continue to increase.
Because the high number of dedicated CPUs needed for provisioning the service,
to meet the demand from a fraction of these radiotherapy centres, implies access to
big dedicated server farms to guarantee the quality of service. However, the proposed
architecture can clearly benefit from the utility computing model as Cloud. Using
similar methods for CPU renting as the proposed in this Business Experiment, it
would be possible to add elasticity to the service and avoid the unwanted overload
which will decrease the confidence in the platform. For example, a full set of virtual
machines can be deployed dynamically and integrated in a local scheduler, using for
example the OpenNebula framework (Sotomayor et al. 2008).
Other important upgrades of the platform in the future are the control of the
Service Level Agreements. Currently we have implemented a model where the
contracts with the providers will indicate the limits of the CPU provision, leaving
for the on-line negotiation the selection of the real values (because, for example, the
number of guaranteed CPUs can change with time. Having a high number of guar-
anteed resources will increase the price substantially). The platform must include
mechanisms to control the agreements. We are studying the integration of other
SLA components such as the SLA Evaluation and Monitoring that allows the plat-
form to enforce agreements. Also, there is room for improvement in the security,
adding capabilities to keep encrypted patient's images (or the data extracted from
