Gird computing example

The problem is based on (though not necessarily resolved in the same way as) a real project: the SETI project. Thus, in order to go deeper into the real solution and into the problems that actually took place in said project, it is recommended that students read the references that can be found on http://setiathome.berkeley.edu/ and also on the website of open-source programs for volunteer computer grids at http://boinc.berkeley.edu/

Proposal of the problem

A corporation, “Non-Profit Corporation (NPC)”, needs to perform an intensive processing of a set of data received from different satellites and antennas. This information arrives as a data stream and is stored in a relational database.

Afterwards, all that information will be divided into information packages that will be subjected to a process of intensive calculation, whose result is, in turn, a data package. The result of the process will be the base of a post-process, consolidating the results that will be stored for future reference of the staff of the corporation.

Figure 1. Contextual diagram of the problem

The amount of information is very high, so several million data packages may be processed in parallel. Moreover, this type of process does not have an industrial profile and a commercial use is not planned, so the profitability of the process is very low, or even inexistent.

As a consequence, the corporation does not have a Budget high enough to hire the computing process of high-performance supercomputing services.

The majority of the project (with the exception of the intensive calculations) will be performed with their own resources, so the system to be modelled is precisely that one that gives a solution to the economic restrictions of the need for intensive calculations.

Figure 2. Contextualization of the system to modelled

Methodology to be employed

The study of the architecture had to be made by following the Attribute-Driven Design (ADD) (Bass, y otros, 2003) and the standard from SEI (ANSI/IEEE 1471, 2000).