[clug] Poor PPPd performance
duncan_roe at acslink.net.au
Mon Mar 12 00:12:45 GMT 2007
Can you reproduce the ARM setup in a PC and run tcpdump on the PPP interfaces?
It can be inportant that software always reads as much as it can from a socket
before writing to it. Otherwise the receiver can get starved of buffers. The
remote sender may time out and resend in this case, or it may simply measure a
lengthy round-trip time.
Cheers ... Duncan.
On Mon, Mar 12, 2007 at 09:42:43AM +1300, Craig Bates wrote:
> Thanks Martijn,
> The underlying protocol is quite simple. It will do resends of PDUs that go
> missing, but on my setup I don't have a radio link, just digital board to
> digital board which is very reliable. It seems more to do with PPP just not
> generating data fast enough or perhaps the Pseudo TTY interface....
> On 3/11/07, Martijn van Oosterhout <kleptog at svana.org> wrote:
> >On Sun, Mar 11, 2007 at 09:48:09PM +1300, Craig Bates wrote:
> >> I'm working on radio link project that runs embedded Linux on an ARM9
> >> platform. Four PPP sessions originated on windows PCs go over the air
> >> and are terminated on the ARM9. The software that takes care of the
> >> air link connects to the 4 PPPd processes via pseudo TTYs. So PPPd
> >> thinks it is talking to a TTY and the other end looks like a file
> >> descriptor that the airlink software reads and writes to. So PPP is
> >> transported by a lower layer of software. Much like PPP is transported
> >> by Ethernet or ATM in DSL. I hope that makes sense...
> >Does the underlying protocol have flow control, like TCP? In that case
> >you should look at:
> >Hope this helps,
> >Martijn van Oosterhout <kleptog at svana.org> http://svana.org/kleptog/
> >> From each according to his ability. To each according to his ability to
> >-----BEGIN PGP SIGNATURE-----
> >Version: GnuPG v1.4.1 (GNU/Linux)
> >-----END PGP SIGNATURE-----
> linux mailing list
> linux at lists.samba.org
Please avoid sending me Word or PowerPoint attachments.
More information about the linux