Return to BSD News archive
Xref: sserve comp.os.vxworks:3324 comp.unix.bsd:14085 comp.unix.questions:50122
Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!msuinfo!uchinews!daffy!uwvax!astroatc!vidiot!ftms!brown
From: brown@ftms.UUCP (Vidiot)
Newsgroups: comp.os.vxworks,comp.unix.bsd,comp.unix.questions
Subject: Unix Socket Problem - please help
Message-ID: <2234@ftms.UUCP>
Date: 2 Jun 94 18:01:13 GMT
Organization: Vidiot's Other Hangout
Lines: 38
We have recently run into a problem with sockets that we have been unable
to solve so far.
We have an instrument with data being collected on a VME-based computer
(running VxWorks), which then transfers the data to a Sun workstation.
There is a process on the VME computer which writes to a socket. The
other end of the socket is read by a process on the Sun, which then writes
the data to a file.
The problem is that if we are collecting data rapidly, we seem to fill up
the socket buffers on the Sun side. Once that happens, we have the following
problems:
1. The performance on the Sun side degrades drastically (CPU usage goes to
100%, and the process reading from the socket only gets a few bytes
at a time).
2. Somewhere along the line, four or eight garbage bytes get inserted into
the data stream for each data set.
We have tried everything we can think of to try to keep the socket buffers
from getting filled up. The only thing that has worked is to add fixed time
delays in the VxWorks process before writing each data set, which is
unacceptable.
Has anyone else seen anything like this, and does anyone have any suggestions?
Please e-mail your suggestions to me instead of posting ('cause if you do I
will miss the reply).
Thanks in advance.
MB
--
e-mail: brown@wi.extrel.com or ftms!brown%astroatc.UUCP@cs.wisc.edu
phone: (608) 273-8262 fax: (608) 273-8719 voice-mail: (800) 426-6488 ext 8293
System Administrator - Extrel FTMS - Madison WI. [An era has come to an end.]
[The need for igniting the midnight petroleum has come to a close as well.]