home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.software-eng
- Path: sparky!uunet!cs.utexas.edu!milano!cactus.org!wiley
- From: wiley@cactus.org (Mark S. Wiley)
- Subject: Re: Testing Parallel Components
- Message-ID: <1992Aug21.213902.8000@cactus.org>
- Organization: Capital Area Central Texas UNIX Society, Austin, Tx
- References: <1992Aug21.160553.22273@evb.com>
- Date: Fri, 21 Aug 1992 21:39:02 GMT
- Lines: 21
-
- In article <1992Aug21.160553.22273@evb.com> brandon@evb.com (Brandon) writes:
- >I'm curious to hear what people are doing to test parallel software
- >components. Any information would be appreviated.
- > Brandon Goldfedder (brandon@evb.com)
- >
-
- It depends on what type of parallel software you are testing. When I
- was testing Sequent's symmetric multiprocessor OS (Dynix/ptx) I wrote a
- parallel test driver to run 100's to 1,000's of test processes in
- parallel. This tends (-: to put it mildly :-) to stress the operating
- system and expose previously unknown race conditions in the OS.
-
- Since you can use standard functional tests running in parallel, the
- investment (writing the parallel test driver) is relativly small compared
- to the amount of insight you gain into how your OS runs in a stressful
- environment.
-
-
- --
- Mark S. Wiley
- mark.wiley@cactus.org
-