* PPP and buffer sizes
@ 2010-01-14 22:19 Kristian Evensen
0 siblings, 0 replies; only message in thread
From: Kristian Evensen @ 2010-01-14 22:19 UTC (permalink / raw)
To: linux-ppp
Hello,
I am trying to debug a performance issue I am experiencing with
L2TP-tunnels (created using xl2tpd). When I measure the UDP-performance
of different desired throughputs using iperf together with the tunnel, I
see that the throughput stabilizes at a certain speed while the packet
loss increases. When I measure the performance using TCP together with
the tunnel, I see a very high number of packet losses, forcing
congestion control and resulting in a significantly lower throughput
than without tunneling.
I guess this is caused by buffers being filled up and packets are
dropped, however, I have not been able to figure out which buffer or
exactly why this happens. If I remove the tunnels and measure the
performance, the maximum supported throughput by the network is reached,
so both the UDP and TCP send and receive buffers are big enough. I have
also measured the processing delay caused by the L2TP-software and that
is at most 50 microseconds, which should not affect the results.
L2TP passes packets to PPP after they have been decapsulated and, as far
as I have understood, PPP + IP layer processing (which is on microsecond
scale) is the last step before the packet is sent to the application
layer. My question is therefore, is there anything in PPP that can
explain what I am experiencing?
Thanks in advance for any help,
Kristian
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2010-01-14 22:19 UTC | newest]
Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2010-01-14 22:19 PPP and buffer sizes Kristian Evensen
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).