A Quick Note about Cloud Computing
I was talking to several people at the Cyber Leap Year Summit about how we have decades of research in computing that too many current researchers fail to look at because it was never put on line. We have all noticed the disturbing trend that too many papers submitted for publication do not reference anything before about 2000 -- perhaps because so much of that early work has not been indexed in search engines?
I mentioned that I had seen papers a few years back where the authors had implied that they had invented virtualization, despite the idea going back decades; at least the Wikipedia entry seems to avoid that mistake.
Someone jokingly mentioned that at least a few things were new, such as cloud computing.
Not so fast.
Some Cloud Computing is really nothing more than SaaS on virtualized platforms. That isn't new.
However, one view of Cloud Computing is that it provides seamless processing and storage somewhere on the net, where you don't have to know where it is stored, where it makes use of multiple platforms for performance and storage, and you don't need to worry about individual machine failures because the rest of the system continues forward.
Interestingly, that was precisely the goal of the distributed OS project where I did my PhD dissertation. I wrote the first prototype distributed OS kernel for the system. The name of the project? CLOUDS. The year? 1986 was when I defended, but the name was coined in 1984. (Cf. a summary article written in 1991.)
My kernel had virtual memory, process creation/deletion, object stores, capabilities, and a built-in debugger (that one could invoke after a crash -- no blue screen, simply the console message of "Shut 'er down Scotty, she's suckin mud agin.") I demonstrated it creating objects and invoking methods (actions) on them across the network on other machines, among other things. Three later PhD dissertations relied on it, as did at least 2 MS theses.
(Oh, and I wrote most of the code in VAX assembler language and it all ran on the bare hardware. I debugged it by stepping through memory, and found some hardware bugs in the process. I was a real programmer back then: i have programmed machine code on six architectures, and in over 25 other high-level languages. But I digress...)
My dissertation is not very good; I would not accept it from one of my students now, and do not recommend anyone read it. But circumstances were such that I didn't actually have an advisor for a big chunk of my research work, and the committee wanted to get me out. I never got a publication from the dissertation work, either. In retrospect, I'm not sure that was the best course of action, but I seem to have turned out more or less okay otherwise.
Bonus item: The first Ph.D. from the group, based on an earlier attempt at the kernel, was Jim Allchin. But don't blame the Clouds group for Windows!
Bonus item: Only about 4 people ever knew, but "Clouds" was an acronym. We liked the imagery because if you combined two clouds, you simply ended up with a cloud. Up close, you couldn't tell where the boundaries of a Cloud were. And if you took some away from a cloud, you still had a cloud. Great, huh? I'm going to reveal the acronym here: Coalescing Local Objects Under Distributed Supervision. We needed the acronym for the proposal to the funding agencies, but for obvious reasons, we never referred to it as anything other than Clouds. The acronym was coined by Bill Thibault.
Bonus item: The Clouds kernel was the third OS I had written, and the final one. The second one was also in assembly language and some custom microcode for the PR1ME 500 &750 series computers (of which Georgia Tech had five). I taught a class around machine architecture and writing an OS at Georgia Tech while a grad student. I'd love to hear from anyone who took the class.
Bonus item: Although my research work quickly moved into other areas of computing, I stayed with OS long enough to help start and chair (with George Leach) the (first WEBDMS; 1989 and) SEBDMS (1991, 1992) conferences. These later evolved into the OSDI conferences -- which I have never attended. It is unlikely that many people remember this connection.
A few people still remember me for that OS work. Others know me for the work I did in mutation testing, and yet others for the work in dynamic slicing and backtracking for debugging. That was all before I started work in security and forensics. I'm to blame for more than many people know -- and I'm not telling about the rest.
But next time someone tries to tell you about their latest "new" idea, you might check with some of us
older more seasoned computing folk, first, and let us reminisce about the good old days.
on Sunday, September 20, 2009 at 06:40 AM