blob: f8037027be1708bdce0487c95a00976bf217327e [file] [log] [blame]
`stress' TODO -- development roadmap
Copyright (C) 2005 Amos Waterland.
See the end for copying conditions.
Please send `stress' bug reports, questions, and suggestions to
<apw@rossby.metr.ou.edu>.
Build easily without autoconf
* Define VERSION, among other things.
Use a concept instead of an option
* Instead of explicit options like --vm-bytes, get rid of all --vm
options but one: --vm-pattern=webserver,database,raytracer etc.
Barrier instead of backoff
* If you fork a sufficiently high number of children, the system will
become so overloaded that new children do not ever get a chance to
even start. It is arguable whether stress should bother to deal
with this at all, but at present it does so by giving each child a
backoff timer. Each child immediately goes to sleep upon entry,
and wakes up only after the backoff factor has expired. This
allows the kernel to fork an absurd number of processes. If we
decide this is a behavior we wish to preserve, we should move to a
model where each child just waits for a signal from the parent,
which it can send to its whole process group effeciently once it
finishes forking all the requested children. This would allow us
to get rid of the `--backoff' option and reduce the associated code
complexity.
----------------------------------------------------------------------
Copyright information:
Copyright (C) 2005 Amos Waterland
Permission is granted to anyone to make or distribute verbatim copies
of this document as received, in any medium, provided that the
copyright notice and this permission notice are preserved,
thus giving the recipient permission to redistribute in turn.
Permission is granted to distribute modified versions
of this document, or of portions of it,
under the above conditions, provided also that they
carry prominent notices stating who last changed them.