|
A DotNet Grid computer for FSCET-SENAC The project The idea to start this project began in 2002 due to a collaboration with the Ludwig Institute for Cancer Research, branch of Sao Paulo. The collaboration is to help analysing microarray data, a task that poses huge computation problems, for instance: finding molecular classifiers of few genes to predict cancer, or assessing the confidence of statistical tests (that may require the use of a very known statistical technique called bootstrap). The typical timeframe to complete a task in a good computer is measured in weeks. At that time, Roberto Hirata Jr. was working for SENAC College of Computer Science and Technology and he had under his supervision two undergraduate students: Paula Akemi Nishimoto and Rodrigo Assirati Dias. Besides that, he had for this grid project the collaboration of Lucas A. M. Perin (formerly Lucas A. Meyer dos Santos), a BSc in Computer Science (IME-USP) and MS expert. The Servico Nacional de Aprendizagem Comercial (SENAC) is a private non-profit education institution that has activities spread all over Brazil. The institution that has more than 6000 personal computers in educational laboratories just in Sao Paulo and a small fraction of them, approximatelly, 200 were in seven didactical labs in the SENAC College of Computer Science and Technology. The SENAC College of Computer Science and Technology (FSCET) is one of the its units. The FSCET has been created in 1998 as an extension to the Center of Education in Informatics (CEI-SENAC Sao Paulo). CEI offers a large and varied amount of computer training programs. Besides that, it is a Microsoft Solution Provider and offers one of the most complete training programs on Microsoft software in Brazil. Due to this historical background, one of the boundary conditions we had was that most of the SENAC's computers run Windows, so should the grid. A suitable platform to achieve this task is naturally dotNet. The initial idea was to build all the grid software but then we found Alchemi. This was of great help because we could concentrate on other problems. The second important boundary condition was that the software had to run without affecting the normal life of the laboratories. The natural way to do this it to make it a service for Windows. Besides that, the service should be easy to install, start or stop. The third boundary condition was that it had to connect with other softwares, like R (www.r-project.org), for instance. In particular, to start a thread to run R we had to use COM technology on dotNet. To do this, the software was adapted to use RCOM, a package contributed to R. Results The implementation, now being released to community, is one of the results we planned to the project. One paper is being prepared about this experience and another on the economical aspects of having a grid computer for an institution like SENAC. There are several interesting economic models that can be used for reducing the cost of ownership for the institution. Another result of the project will be a technical article about using this grid model to maintain and manage R installations and package distributions along all workstations of a network. Future work Roberto Hirata Jr. is working for the University of Sao Paulo since the second semester of 2004 and he is supervising Rodrigo Assirati Dias, who started pursuing his master's degree this year. Their plan is to continue the work in collaboration with SENAC and finish the experiments with the new grid of 900 machines the institution has on its new campus (that would put SENAC in the top 500 super-computer machines of the world). |