NOS & heavy Ethernet traffic
- To: mikebw@IDS.JVNC.net (Mike Bilow email@example.com)
- Subject: NOS & heavy Ethernet traffic
- From: Marcel Wiget <firstname.lastname@example.org>
- Date: Sat, 11 Jan 92 16:58:29 MET
> This is really strange. Did you try simply increasing the "mem nibufs"
> parameter? Or "mem ibufsize"? It is very interesting. -- Mike
Yes Mike. Increasing the mem variables doesn't help. It is a hard
test with 'spray', because the "receiver host" hasn't even a chance
to send something else most of the time. And NOS doesn't do more than
to try to read all packets.
The routine in NOS, refiq(), has more chances to refill the ibufs
than other processes and so all memory is occupied by ibufs.
I think that per packet will be a complete ibuf as sized by ibufsize
used. That meens alwayis 2048 bytes per packet. In this case ibufsize
should be extreamly high.
But with my previous sent patch I couldn't any more crash NOS with
'spray' even with additional activities via telnet and ftp.
vy 73 de Marcel, HB9RWM