tc, tbf and accuracy

Linux Advanced Routing and Traffic Control

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi all,

   I've tried to look for this info in previous messages but had no luck.

I've a 2.4.20 kernel on an Intel arch., all TC is configured and it seems to work. Anyway, I've had some troubles in configuring TBF via tc:

1. when I configure burst or minburst to be, say, 1500b, when I peek the configuration with "tc -s qdisc" the answer is that burst/minburst is equal to 1499. Moreover, any other value I try to configure in "b" unit is always decremented by 1 (3000->2999, and so on). Any clue about that?

2. I also have some strange behaviours about latency configuration - it is always far from what I request. If, for example, I try to set 500ms, the most probably I get something around 620ms. Note that if I configure the "limit" parameter instead of "latency", I can manage to reach values closer to what I need (e.g. 500ms). Is there something I'm missing about latency configuration?

3. I've tried to configure peakrate/minburst, but I had to struggle a bit to have them work. In particular, when I tried to set minburst equal to interface MTU (100 bytes, ethernet) as suggested in the manpage I did not even manage to set up an HTTP session. I had to configure a minburst equal to around 12000 to have some traffic pass arrive to the remote client. Is the manpage wrong?

4. What I really need is to have a controlled bit rate for a unique HTTP session from my PC (i.e. no other traffic during the tests I want to perform). I was thus wondering what does it mean that there is a "drop" in packets that are generated from my own machine. I thought that there could be two basic models:

4.1 write-wait model, in which the rate is limited directly towards the application that issues some "write" on the socket file descriptor, thus limiting its rate - or

4.2 tcp-wait model, in which the application talks to TCP very fast, and some packets from TCP are actually dropped if TCP tries to send them too fast.

What model is actually implemented?


Thank to you all for any bit of attention,

	Flavio.
_______________________________________________
LARTC mailing list / LARTC@xxxxxxxxxxxxxxx
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

[Index of Archives]     [LARTC Home Page]     [Netfilter]     [Netfilter Development]     [Network Development]     [Bugtraq]     [GCC Help]     [Yosemite News]     [Linux Kernel]     [Fedora Users]
  Powered by Linux