simulation@uflorida.cis.ufl.edu (Moderator: Paul Fishwick) (09/16/88)
Volume: 4, Issue: 10, Fri Sep 16 09:13:16 EDT 1988 +----------------+ | TODAY'S TOPICS | +----------------+ (1) ISPS under UNIX... (2) Chaotic System Behavior Moderator: Paul Fishwick, Univ. of Florida Send topical mail to: simulation@uflorida.cis.ufl.edu ----------------------------------------------------------------------------- Date: Tue, 13 Sep 88 09:30:37 cdt From: reed@m.cs.uiuc.edu (Daniel A. Reed) Office: 136 Digital Computer Laboratory Telephone-Number: (217) 333-3807 Reply-To: reed@m.cs.uiuc.edu To: simulation@uflorida.cis.ufl.edu Subject: ISPS under UNIX Mats Bjorkman from Uppsala recently asked this question. I too am interesting in possible answers... Dan Reed Department of Computer Science University of Illinois reed@a.cs.uiuc.edu ------------------------------ Date: Fri, 16 Sep 88 09:03:16 EDT From: Paul Fishwick <fishwick@fish.cis.ufl.edu> To: simulation@bikini [[This article appeared in comp.theory.dynamic-sys and it contains some relevant points concerning modeling and simulation. So, I'm forwarding it to our list. Please feel free to post other simulation- related articles that you may find on the net. Any followups to this article should probably cross-post to sci.math and comp.theory.dynamic-sys -paf]] >From: Kevin_P_McCarty@cup.portal.com Newsgroups: comp.theory.dynamic-sys,sci.math Subject: chaotic computer networks Date: 15 Sep 88 10:43:30 GMT Organization: The Portal System (TM) Xref: uflorida comp.theory.dynamic-sys:119 sci.math:2835 XPortal-User-Id: 1.1001.2214 There was an article in the Sunday New York Times (Sept. 11) (The Week in Review, p. 6) on chaotic behavior in computer networks. This is an interesting phenomenon in itself; anyone with further information or references is encouraged to share it. Also, Alan Perlis, a Yale computer scientist, is quoted in the article. His remarks puzzle me; they seem to be quoted out of context (see my remarks below). Excerpts from the article: "As computer systems continue to grow more complex, scientists are finding it increasingly difficult to accurately predict their behavior." [...] "Recently, for example, computer designers at TRW, the weapons manufacturer, were surprised to find that a large computer network they had strung together in Europe was exhibiting strange, unpredictable behavior. On close examination the engineers discovered nothing wrong with the design of the system, which linked together hundreds of computers as part of a military data communications network. "The engineers now suspect that they were confronted with the mathematical concept called chaos, a natural phenomenon that leads to turbulence in rapidly moving water or in the atmosphere. [...stuff about determinism vs. controllability...] "But if a computer network is, like the weather, subject to chaos, then there will always be a danger of constructing systems that elude the control of their inventors". [...(we didn't need chaos theory to tell us this)...] "A group of experimenters at Xerox Palo Alto Research Center in California recently conducted a series of experiments that indicate that large aggregations of connected computers can exhibit unpredictably wild oscillations and unstable behavior. The research offers new insights into the behavior of the immense data networks that are blanketing the planet. [(does anyone have further information on this line of research?)] [...stuff about computational ecology and distributed computing...] Here's the part that puzzles me: "Alan Perlis, a Yale computer scientist, suggests that the chaos problem lies in the inevitable disparity between the real world and the models used to simulate it. Even the finest computer simulation is only an approximation. At some point that cannot be determined in advance, the discrepancies between reality and the computer's simplified world view will lead to a chaotic breakdown." Perlis seems to be talking about an entirely different topic. There are no models in a computer network, it's a real thing with real behavior that is being observed 'as it happens', not being simulated. I.e., it's a bona fide 'natural' phenomenon. In addition, chaos is a structurally stable behavior of systems, so there is no reason to believe that such stuff only happens in computer simulations and not in reality. My initial impression of the statement was that occurrence of chaos was due to inherent innacuracies in modeling. While it's true that the occurrence of chaotic behavior may make a model's quantitative predictions worthless, a prediction that chaotic behavior will occur should be reliable. Perhaps Mr. Perlis meant to criticize the experimental methods of the Xerox scientists, as not reflective of the actual situation seen by TRW; this may be so-- it's not clear from the article. On the other hand, he may have been talking about network traffic models which suggest control strategies which are theoretically stable, but not when implemented in practice. Are there any network traffic models which have been demonstrated to exhibit chaotic behavior? Kevin McCarty ------------------------------ +--------------------------+ | END OF SIMULATION DIGEST | +--------------------------+