John T. Robinson
Note: this web page will be going away in October 2009; after that
I will be using my LinkedIn
public profile as a home page.
My general area of interest is the design and analysis
of algorithms and computer systems, at both the software
and hardware levels. I have over thirty years experience in
this area, in fields including database and operating systems,
parallel and distributed algorithms and systems,
storage hierarchies and systems, buffer and cache management,
data visualization, performance analysis, compression technology,
and hardware design verification.
Some of my more well-known work includes optimistic methods
for concurrency control, K-D-B-trees, frequency based replacement (FBR)
methods for cache and buffer management, analysis of limitations
of concurrency, concurrency control methods for high contention
environments, and compressed memory system architectures (part
of the basis for IBM's MXT product).
I was a Research Staff Member at IBM for over 23 years
until I left (as the result of a "resource action") in 2005.
It was an interesting experience (the first time that what I now
think of as the type 2 management style had such a
direct impact on me), but not particularly fun.
Since then, I have finished some in-progress work (including some
part time work for the IBM IP Law department on
patents - they called me after I left, and yes, they paid me),
and have started doing independent
research in the general area of analysis of algorithms (having
made good progress in a couple of new areas).
Currently I am seeking a new position, either academic or in
industry, and also investigating small business ventures. Here is my
resume.
After leaving IBM Research, I wrote this short
essay on research management culture. Thanks to feedback from a couple of friends,
this revised version is somewhat more balanced and rational
(at the cost, possibly, of the humor becoming more obscure).
Here is a list of
useful and/or interesting sites,
which also partially reflects some of my other current interests.
Is the following true? Let f(x) be a solution of the differential equation
f''(x)f(x)=f'(x), and let p(n) be the n'th prime number. Then asymptotically p(n)~=f(n).
- John