This post was brought on by a seminar I was in yesterday; and as I was sitting there, several things struck me; especially the fact how no-one else in the room really thought about the things I was thinking about.
The talk was given by an ITaP person – I do not remember his name (nor would I like to disclose it, if I did, for after all, what i have to say, has nothing to do with that person anyways, nor indeed would I like to violate someone’s privacy). ITaP stand for Information Technology at Purdue – at Purdue University, all computers are administered and maintained by this department. From time to time they also build supercomputers for the University, among other things. Purdue after all, is an engineering school, and computing is an essential part of what Purdue does.
And now onto what caught my attention. The speaker mentioned that Purdue University was one of the first Universities to build its own supercomputer – some Internet research reveals this to be the CDC 6500. And in the very next slide, he said that Purdue built the Nation’s fastest supercomputer again in 2013 – this was the Conte.
More info about computing resources at Purdue University:
But is it all business as usual? Let us probe a little bit further.
In the 1960s, leave alone a super computer, even assembling a computer was not an easy task. First, there was a variety of CPUs you could choose from – IBM, Motorola, and a number of other companies whose names have today been forgotten – like CDC and VAX – made CPUs. Each CPU was almost a custom design back then.
And it did not end there – for next came the RAM, the Hard Disk (More likely a Tape Drive or several of them) – all of these had to be handpicked too. And each of these components did not have the reliability or the speeds that we expect of them today – no, they failed much much more often, sometimes lasting less than a day. Speeds were slower too – and capacities were much smaller. A storage of 1 MB was a dream back then.
But still further, was the design of the system, and this is what I want to really talk about. These days, you can go out and buy a hard disk, and simply plug into your computer. Same for the RAM, the graphics card, and even, to some extent, the CPU. What enables this?
It is the motherboard. Each computer has one, and it is on this component that all other components sit on – the CPU, the graphics card, the RAM, the Hard Disk, everything connects to the Motherboard. The Motherboard is like a circuit – in fact, it is very literally a circuit – a map of how the different components will connect together, and talk and interact to each other.
You probably have not heard of the Motherboard though. Most likely, even if you assembled your PC yourself, you, chose the CPU, the graphics card, possibly the RAM with the greatest of care, and when it came to the Motherboard, you just chose whatever was on offer – possibly something from ASUS if you wanted to splurge, ASUSTek or GIGABYTE, if you were mainstream, or BIOSTAR if you were really cheap. Possibly Intel, if you just bought a bundle.
But, and this is what I’m onto – there were no motherboard – let alone motherboard manufacturers back in 1969. No – you actually had to build the thousands of circuits on a Motherboard by hand – and then you had to check them for correctness by hand. And then you had to actually connect all your components in that exact same design – also by hand. Building a computer in the 1960s was not easy – by any means.
And there was more – and this is the part on innovation.
When you built it all by hand, you had the freedom to do it your way. So the CDC 6500 at Purdue was a unique creature. It must have been designed by Professors at Purdue and Graduate Students with great patience, it must have been built, again with great labor, and when it was finally built, there was no other computer in the world that had the exact same configuration. The CDC 6500 represented a unique human effort, that was uniquely a Purdue product – there was nothing quite like it anywhere else.
And it goes even further.
For on the Hardware, sits the software stack.
Back in 1969, there was nothing like UNIX, or even something resembling an OS. So again, Professors and Grad Students at Purdue must have built an Operating System for their beloved CDC 6500 – something possibly given to them by CDC or IBM, but something that they also must have subsequently modified to fit their own unique needs.
At the time, even languages were not really standardized – so that again, Purdue Implementations of Scheme, Lisp, FORTRAN – and who knows what other languages – must have been loaded onto that CDC 6500.
Again, there was innovation – there was a uniqueness – there was what one can call a Uniquely Purdue Product – the software stack that sat on this unique piece of hardware, also designed by Purdue.
There was innovation, there was uniqueness.
The Conte, the supercomputer that came online in 2013, is none of these things.
Supercomputers today, are made of exactly the same components that you find in your own desktop computer today – except one uses a larger number of them.
Thus, a supercomputer would have instead of 1 CPU, upwards of a 100. Instead of 1 hard drive, it would have a thousand. Instead of 1 RAM chip, it would have hundreds.
And so, a supercomputer of 2013 is a commodity design. If you have enough money, you can build one in your own backyard – there is no innovation here, no new technology.
And the software stack too – they mentioned HADOOP and MongoDB – both these technologies again arose from the Industry. Here too, we have a blandness.
The “Uniquely Purdue” aspect is totally missing.
It is not that we are against technological advances – but what we are onto is the aspect of innovation. There is no innovation in what Purdue did in 2013 vis-a-vis what it accomplished back in 1969.
Yes, even in 2013, you can develop new CPU designs – but it is happening in the Industry. Whether it is the FPGA companies (Xilinx, Altera, etc) or the unique Micron Automata CPU – this technological development is happening almost entirely in the Industry.
Yes, Purdue did build one of the World’s fastest supercomputers in 1969, as it did in 2013, but did Purdue innovate and invent as much it did in 1969 as it did in 2013? Did any University?
These are the kinds of questions those in Universities should be asking themselves. It is on these questions that their future and the future of the places they work in depends on.