Blue Gene/L - supermachine?

[baudrunner]

IBM has developed a new supercomputer, reputed to be the fastest computer on the face of the planet. It is capable of astounding computational speed. This computer utilizes 131,000 Power PC processors massively paralleled in a hardware setup that would crowd your living space.

Is this design overkill? Before you ridicule this idea, have a read. And then reconsider your personal perception of reality.

I used to program computers for a living. I used a hugely popular sophisticated commercial programming environment with a huge library of functions which were all written in C and assembly language. I also never used the supplied meta functons which provided the high level data browse and search routines, because I discovered that writing my own routines that did the same thing produced much faster and way more efficient, not to mention considerably smaller, data display and manipulation functions, especially as the data sets grew larger and larger.

Complex computer code embedded in large modules are difficult for any programmer to grasp if they have not written the code themselves, and over-documenting the code only adds to the aggravation of that task.

I am of the opinion that IBM likes simple, stupid, linear code, and no matter how intelligent and efficient a modular program is, and no matter how clean and logical the recursion of called functions is, they will invariably reject a clever program for being too smart. I have personal evidence of that fact. They will call the computer code to be "convoluted", or "spaghetti code". This, in spite of the fact that the compiled program may be completely bug free and really very good.

You see, the bottom line is MONEY.

In this world of increasing microprocessor circuit density and ultra-miniaturization there is a fear among large mainframe manufacturers that they are becoming dinosaurs. Their profit strategy lies in hugely over-priced ultra-expensive and very large and imposing metallic boxes that fill a room. If IBM optimized its engineering to produce very fast and very small boxes that fit on a desktop and utilized intelligent generic operating systems and clever software they would soon go out of business.

There once was an intelligent physicist by the name of Williams who took thirteen years to build a computer of his own design because nothing existed at the time to do the work that he needed to do, which was to calculate the probabilities of sub-nuclear particle properties and characteristics in order to prove that each and every particle in existence had its very own unique place in the cosmic scheme of things. By the time he was finished, he could have accomplished the same task by parallelling a couple of Pentium 1's, which were the cutting edge in PC's at the time.

Which begs the question, why does IBM still build them so big? Is there a software problem? Or is it a deliberate strategy to reap huge profits on the sale of appartment sized hardware? I think the latter, myself.

Fact is, we don't need to parallel 131,000 Power PC processors to do the tasks that researchers are demanding of them. What we need is intelligent software, but what I actually see happening is that there are very, very few intelligent programmers out there who are allowed to work outside of a large group in a team environment in a colossal institution. That is one of the reasons that Windows is hundreds and hundreds of megabytes large. Believe me, that is not a good thing.

[Aiya-Oba]

Hi baudrunner:
I am taken by your little quote: "There is nothing permanet except change." There is nothing more desirable to the mind than understanding.That permanence is change, is most revealing.
-Aiya-Oba

[harmonygirl]

Baud, welcome to the way I view the world. Power and money dictate that things are not done efficiently, or at optimal capacity but for profit (or other benefit) maximization. Look at breast cancer. By all accounts, given fundraising, media attention, celebrity endorsement, and dedicated researchers, logic dictates that there should have been at least an attempt at a cure by now (if not the cure itself). There is, however, no money in curing. There is lots of money in treatment, tho. The unfortunate thing, is that very few people critically analyse the underlying structures of our society and then have the bravery to speak out or do something about it. Maybe you should write a book.

[baudrunner]

Thanks, HG, and Aiya, for your replies.

It has often occurred to me, and continues to occur to me, that it is quite possible that it is the penny which prevents humanity from going to the next level. I have often expressed the desire to rid the world of prices that end in $n.99, which exist just to make the purchase look more attractive. It is indeed true that there appears to be nothing permanent except ¢hange, especially in the copper sense/¢ents. Sometimes I feel compelled to start a movement to rid the world of the penny.

I trust that you have appreciated the stubborn paradox in my signature. It will not change so long as I don't.

And, insomuch as life goes on...,