Today's Topics:
Slack in CS
----------------------------------------------------------------------
Date: Mon, 18 Nov 91 09:17 MST
From: AHEWARD@ccit.arizona.edu
Subject:
To: Subgenius@mc.lcs.mit.edu
Message-id: <E2EFEED6895F4042A3@ccit.arizona.edu>
About a week ago someone was asking for information about the _Fortean
Times_. It appears that no one else is going to appease this poor person's
pleas for help so I guess I'll do it.
The _Fortean Times_ is a quarterly "Journal of Strange Phenomena"
out of England. A good way to describe it would be to call it "The Skeptical
Enquirer," because it deals with the same sorts of crazy claims as the "Weekly
World News," but in a skeptical light.
The latest issue has a good series of articles about claims of Satanic
child abuse in England. It also has a section entitled "Not Their Day" with
a bunch of news clippings from around the world. For example:
"A madrid postman stole a parcel from work, took it home and opened it.
Unfortunately it was a letter bomb and it seriously injured his wife. Police
charged him with theft. _Sun_ 19 Dec 1990."
Single back issues cost $5, and a four issue subscription is $16.
Send checks, etc., to Fortean Times, 20 Paul St., Frome, Somerset BA11 1DX.
------------------------------
Message-Id: <9111190119.AA29120@mahler>
To: subgenius@mc.lcs.mit.edu
Subject: Slack in CS
Date: Mon, 18 Nov 91 20:19:07 -0500
From: Michael Travers <mt@media.mit.edu>
Fault tolerance is what it's all about, after all...
Date: Mon, 18 Nov 91 11:52:18 EST
From: menke@das.harvard.edu (Baiba Menke)
Subject: CRCT Colloquium Series Presents
COLLOQUIUM SERIES PRESENTS
Using Slack in Dynamic Fault Tolerance
Yonatan Aumann
Computer Science Department
The Hebrew University of Jerusalem
Thursday, November 21, 1991
4 PM
Aiken Computation Laboratory 101
(Tea at 3:30 pm Aiken Main Lobby)
Abstract:
The introduction of parallel computing has also introduced,
with it, greater possibilities and higher probabilities for
hardware failures, with the probabilities growing exponentially
with the size of the system. Massively parallel systems must,
therefore, be equipped with methods to tolerate these faults without
severely degrading the systems' performance.
Recently, Valiant has shown that with the introduction
of a logarithmic "slack", efficient parallel computation can
be carried out on practical architectures. In our work we examine
the usage of slack in constructing algorithms for efficient fault
tolerant computing in the presence of dynamic failures. We show
that with the introduction of slack the algorithms become both
simpler and stronger than those previously known. Specifically,
we show that if a K processor PRAM is to be emulated by an N node
hypercube, with K/N=Omega(log N), then the emulation can be carried
out with optimal slowdown (up to a constant factor), even in the
presence of dynamic failures.
We then briefly consider the case of other networks, chiefly
the two dimensional array. Finally we note on some related open
problems.
Joint work with Michael Ben-Or.
Host: Professor Victor J. Milenkovic
------- End of Forwarded Message
------------------------------
End of Subgenius Digest
******************************
