Navigation
© The HOISe-NM Consortium 1997
 |  mail the editor |
"When the entire computer is on a chip implementations will be cheaper and easier to program"
Barcelona, 24-03-1997 Professor Veljko Milutinovic comes from the University of Belgrade and he was pleased to stay in Barcelona for a few days, not only to talk about advanced computer architecture during the courses organized by the Interdepartment Commission for Research and Technology Innovation (CIRIT) this month at the Polytechnical University of Catalonia (UPC), but also to learn much more about a subject in which he considers himself as a "very important research place" in supercomputing.
Primeur Do you consider the Polytechnical University of Catalonia as a very important research facility...
Veljko Milutinovic Yes, it is, and not only in Europe but worlwide. Computing architecture is one of the fields where American universities are the elite. And there are very few places elsewhere which are competitive and which do this research which is at the leading edge. This is one of the few places outside America which is at the best level, especially the the Computer Architecture Department of the Polytechnical University of Catalonia. I came here to deliver a Tesla Foundation medal (Tesla was one of the world's ten most important researchers in Electrical Engineering [according to IEEE Spectrum]. He came from Serbia) which recognises outstanding people in the field of Tesla's research, which was power engineering. Tesla medals are given each year to outstanding institutions and individualsin several fields, among which is computer architecture. This year, the Computer Architecture Department of UPC received an award and I have had the pleasure to deliver the medal to the head of the department, Mateo Valero. In the University of Belgrade, we reference a lot of the work which is done here. And we have to underline that the fact that any university outside America which has made such a contribution to Computing Architecture gives great motivation to other universities all around Europe.
PrimeurWhy do you think Distributed Shared memory systems are competitive?
V.M.It is very important, on the one hand, to have lots of microprocessors working together in a box, and on the other hand, to have something which is easyto program. In fact, there are some machines which are easy to program but limited in their expandibility, so you can expand them only up to about 16 or 32 microprocessors. You also have machines which are easy to expand (to a hundred or thousands of microprocessors), but extremely difficult to program. Distributed Shared Memory systems basically represent a successful marriage between the first and the second options.
Primeur So they are easy to expand as well as easy to program...
V.M Exactly.
Primeur This architecture can be implemented in both hardware and software forms...
V.M.Yes. There are diverse approaches. If you do an implementation in software, you should do it on the top of off-shelf-platforms, like a message-passing platform, or on top of workstations. Software implementation is quite cheap but slow, the opposite of hardware implementations: they are much faster but the cost is higher.
Primeur You are one of the researchers who believe that in the future there will be a full DSM computer on a chip. What are the advantages of this?
V.M.Well, it is too early to predict the future with accuracy. But, in these systems everything is on the chip. Chips can be produced in large quantities, thus implementations will be less expensive and easier to program.
Primeur And the machines will be smaller...
V.M.Yes, you are right and that is important.
Primeur Small and cheap... Will they be more performant than a PC?
V.M.PCs will be powerful but distributed shared memory systems will be able to do far more for the same money. So it will be useful to have one of them.
Primeur You have been in the States for about a decade and you returned to Serbia when the war broke out. It might have been difficult to do research then...
V.M.All my education was in Serbia, including my Ph.D. I was in the States as a professor of one of the top five Computer Engineering Schools for a decade and then I came back to Serbia in 1990. I did research for the local and international companies, most of them from the USA. The war was a hard experience but it did not affect my work, because during this kind of situation the best you can do is to sit down and do research. You become very quickly used to the environment. The human being is very flexible. You know, when I was a professor in the States I wrote for a major journal from time to time. Last year, for the first time I published a paper each month in IEEE journals, from Serbia! Moreover, in my country, I wrote several books which are used by students worldwide. So, sometimes, irregularities in the environment can cheer up your imagination.