Messages in this thread |  | | Date | Sun, 08 Sep 1996 20:55:04 +0200 | From | "Robert O'Kane" <> | Subject | Net device- can't feed fast enough |
| |
Greeting,
I am using the ISDN-4-Linux package and have a general question,
i486-dx4/100 Linux 2.0.18 Isdn-4-linux 2.0 (i4lutils 2.0) Teles 16.3 cards Net-utils 1.32alpha Libc 5.3.12 gcc-2.7.2 (both host configurations identical)
The I4L package has the ability to bind together multiple channels over 1 "kernel" net device. The slave device is triggered when the master device is "loaded". For I4L, it kicks the slave device in at 7000 cps.
This works fine except I can't get 7000cps through (to?) the device. Using RAW-IP encapsulation and starting an FTP process, I4L tells me it is doing arround 6040 - 6146 cps so the slave doesn't kick in.
I re-compiled the I4L so the slave starts at 6000 cps. This works fine except that I get again, arround 6040 cps TOTAL across the interface. (with ftp "put" and "get" if that makes a difference). I get the same results when starting the transfer from either host.
Is there a way to see where the bottleneck is?
I'm not sure what to describe so....
tcpdump runs but wasn't interesting except it looks like a "bursting" problem (lots of packets with matching ACKs,, then 5-10 bursts of data, then 5 sequential ACKS). Also, all the FTP data packets were signed DF (don't fragment) and P (push), but I think that is normal though (compared to the ethernet ftp session). The same ethernet ftp transfers were in the 400K/sec range.
Any help or thoughts would be appriciated.
Thanks, Bob O'Kane
-- ---------------------------------------------- Unfortunatly, I find language (like politics) necessary for humans. Leider finde Ich Sprache (wie Politik) fuer die Menschen notwendig. ---------------------------------------------- Robert O'Kane OTHERSPACE Founding Member and Linker
URL : http://www.rz.uni-frankfurt.de/~okane email: okane@em.uni-frankfurt.de ----------------------------------------------
|  |