Imprimir Republish


Each one receives a piece of the problem

Parallel computing replaces supercomputer

The administrator Sidney Pio de Campos still remembers the old days. “We couldn’t cope with all the users who had problems with their connections”, says the man in charge of the computer network of the Gleb Wataghin Institute of Physics (IFGW), of the State University of Campinas (Unicamp). The Campinas institute’s system was not just slow. It was precarious and unreliable. The old network structure, for example, had no control panel capable of pinpointing where the origin of a problem was. “You had to go all over the institute, building by building, until finding the spot with the problem and fixing it”. In most cases, the problem would be a loose cable.

The solution started to appear in 1995, when the institute received the first funds from FAPESP to modernize its computer network. The whole structure was replaced by a more modern system, based on optic fibers. Nowadays, optic fiber links all the 14 buildings of Unicamp’s Institute of Physics. The system has over 700 points on the network, but this does not frighten those who are responsible for it. Its total capacity is for more than 1200 points.

An enormous increase in the speed of data transmission has become possible. When the Internet appeared, in the 80’s, the fastest transmission speed was 56 kilobits per second. By the beginning of the 90’s, this had already leapt up to 45 megabits per second. Good networks today can reach speeds of between 100 and 155 megabits per second. The line that links FAPESP to USP is already working at a speed of 1 gigabits a second. The experts say that the day is not far off when the local networks will achieve this same speed.

Scientific publications
For this institute’s director and FAPESP’s president, Carlos Henrique de Brito Cruz, it is still too early to assess the real impact of networks like these on scientific research. “The results will only appear ten years from now, when the growth of scientific publications and theses published may be examined”, he says. “It is still not possible today to perceive this growth. But we can surely state that facilitating the flow of information always increases the speed and the quality of the production of knowledge.”

Marco Aurélio Pinheiro de Lima, from the Department of Quantum Physics, endorses this. “These days, it is the computer system of an institute that determines its creative capacity”, he declares. If the infrastructure is bad, you can be sure, right away, that research will not get very far. The problems are very sophisticated and call for high performance computing. “The network has made connections more stable and the data transport quicker, with a speed of up to 100 megabits a second. The problems of physics today, for example, involve complex calculations and the transfer of a great volume of data. With a good connection, a researcher can use several computers at the same time. The effect is the same as using a supercomputer.

Known as parallel computing, this resource is widely used by physicists at Unicamp. A single problem is split up into several parts, and each part then goes to a CPU. When the calculations are finished, the data comes back to be put together on a single machine, which controls the whole operation until the final result is reached. Made possible by a high quality network, parallel computing has many advantages. In the first place, it spares the institute heavy investments in more sophisticated machines. Secondly, it can be used from any point on the network. “It is common here for a researcher to ask permission to use other users’ machines, when they have spare capacity”, says Professor Pinheiro de Lima. “When a researcher gets a machine, it is automatically connected to the network. If the researcher does not use it full time, he tends to share it with someone who needs it. That is why our equipment is in use 95% of the time, including at the weekends.”

International research
The network has brought other changes to the institute’s day to day. E-mail has become by far the most used means of communication, whether for internal contacts or for the external ones. “If you need a quick reply, sending a message over the network is a better guarantee than using the phone”, comments the president of the institute’s information technology commission and the Professor of the IFGW’s Cosmic Ray Department, José Augusto Chinellato. Professor Chinellato thinks that the network has also made it possible to take part in important international research, the Auger project. With support from FAPESP, researchers from Unicamp are taking part n this project with scientists from over 20 countries, operating and analyzing the data collected from the Pierre Auger observatory of cosmic rays, in the south of the province of Mendoza, in Argentina.

Without a network such as the one Unicamp has, connected with the ANSP network and the Internet, these scientists could not even dream of taking part in the project. Its objective is to detect, examine, and interpret rare high energy particles that enter the atmosphere, coming in from the space. The scientists’ hope is to obtain more information on the big bang, the great explosion that, according to the most widely accepted theories in Physics, gave birth to the universe. This is a project that involves so much technology and so much money that international cooperation has become a necessity.

Brazil’s contribution alone is to come to US$ 3.5 million, part of this amount invested by FAPESP. “Taking part in projects like Auger would be totally impracticable without a rapid means of data transmission”, comments Professor Chinellato. It was only possible for the IFGW to go in because the institute is up to the project in technological terms.” Chinellato tells he keeps in constant touch with researchers from the United States, Russia, China, Argentina, Greece and other countries, to discuss and exchange information. The data collected by the observatory is sent every day to a database located in Italy.

He was not the only one to be benefited. Part of the work of researcher Douglas Galvão, from the institute’s area of Biophysics, is to pinpoint potentially cancerous molecules and to propose more efficient drugs for their control. This research involves calculations that are only possible with high performance computers. Previously, Galvão would frequently turn to the Cray supercomputer at the Cenapad center, in Rio Grande do Sul.

Using the equipment at a distance, though, made the work difficult. “The situation was critical, since there were many people using the supercomputer”, he recalls. With the network’s capacity for parallel computing, Galvão started doing his calculations in the institute itself. “This helped our research a great deal, he states.

The cases of the Institute of Physics in Campinas are not isolated ones. Research today tends to be more and more interdisciplinary and cooperative. There are many areas where having access to new technologies or not may mean that a group has or does not have the capacity to produce science. “There is today a clear-cut distinction between the countries that have access to information technology and those that do not”, comments Professor Brito Cruz. “That is why the great merit of FAPESP’s infrastructure program was to put São Paulo universities alongside those that have access to this technology”. But he himself warns: “We cannot think that it is all ready. The evolution of this technology is very fast”.