keshav@ernie.berkeley.edu (Srinivasan Keshav) (11/04/87)
I am constructing a synthetic file system workload generator, in the spirit of Satya's file system driver at CMU. The output of the generator will be used by a file system simulator in order to compare file caching policies in a distributed file system. In order to generate file references, I need to look at file sys- tem traces at the level of file operations. Existing traces of file system usage at this level have been done mainly on Unix machines, and traces have been collected at a single node (John Ousterhout, Jim Thompson @ Berkeley, Rick Floyd @ Rochester,Chris Kent @ Purdue, Satya @ CMU, Williamson & Bunt @ Saskatchewan ). I am interested in contacting other researchers with similar in- terests to discuss the parameters in the workload model. I would also like to get in touch with people collecting traces of file operations on non Unix machines (IBM?), or on a network of works- tations. Any replies would be greatly appreciated. S. Keshav (keshav@ernie.berkeley.edu)
cmt@myrias.UUCP (Chris Thomson) (11/05/87)
Collecting this kind of data is very easy on an MTS system. I did a lot of it when I was designing the disk cache for that system. Contact someone at the Computing Center at the University of Michigan, or at the University of Alberta or the University of British Columbia. They could likely send you a tape with a few minutes of live system file activity with a few hundred users logged in. -- Chris Thomson, Myrias Research Corporation alberta!myrias!cmt 200 10328 81 Ave, Edmonton Alberta, Canada 403-432-1616