The file paper.ps contains the CSRI technical report: 

          Implementing Multiprocessor Scheduling Disciplines

                                  by

           	 Eric W. Parsons and Kenneth C. Sevcik


If you have the UNIX uncompress program, get the paper.ps.Z file.
Remember to transfer the file in binary mode.  Uncompress it, and 
print it on a postscript printer.

If you do not have uncompress, get the paper.ps file in ascii mode, 
and print it on a postscript printer.

This report was made available for anonymous ftp by: 

                         eparsons@cs.toronto.edu

Abstract:
  An important issue in multiprogrammed multiprocessor systems is the
  scheduling of parallel jobs, which has led to a considerable amount of
  analytic research in this area.  One of the criticisms frequently made
  about this research, however, is that it is rarely ever implemented and
  even more rarely incorporated into commercial scheduling software.  In
  this paper, we seek to bridge this gap by describing how at least one
  commercial scheduling system, namely Platform Computing's Load Sharing
  Facility, can be extended to support a wide variety of new scheduling
  disciplines.
  
  We then describe the design and implementation of a number of
  multiprocessor scheduling disciplines, each differing considerably in
  terms of the type of preemption that is assumed to be available and in
  terms of the flexibility allowed in allocating processors.  In evaluating
  the performance of these disciplines, we find that preemption can
  significantly reduce overall response times, but that the performance of
  disciplines that must commit to allocations when a job is first activated
  can be significantly affected by transient loads.  We also observe that
  the behaviour of many disciplines over short periods of time can differ
  considerably from that over long periods of time.  Given that users are
  more likely to observe short-term behaviour, future research in this area
  would benefit from studying disciplines at this level.