regarding 'tc' performance

Tang Chin Hooi (hooiye@network2.cs.usm.my)
Mon, 27 Dec 1999 13:41:50 +0800 (MYT)


Hi,
can I ask u is it the performance of tc by using cbq and route classifiers
was not so accurate as what we set? The actual transmission rate normally
will exceed what we have set. Because this was what I concluded after
trying the tc by setting the rate ranged from 100Kb till 9Mb.

Our LAN bandwidth is 10Mb/s and i set the tc by controlling the
transmission rate to 161.142.8.21 from my PC (10.39.11.23).
I tested the performance by ftp the file to the 161.142.8.21 with
different rate setting and got the average of 10 actual
performance results.

Here is the results:

CBQ setting(Kb/s) average actual performance(Kb/s)
100 100
500 517.6
1000 1056
1500 1648
2000 2232
2100 2320
2200 2448
2500 2904
3000 3592
3300 4072
4000 5504
4500 5760
5900 5760
6100 5744
6300 5784
6800 5704
7000 5776
8000 5744
9000 5760

>From the result, we can see the trend that the difference between the CBQ
setting and average actual performance increase as the rate increasing.
The actual performance always exceeds the CBQ setting. Finally, the actual
performance become quite stable because it reachs the
current system performance (best effort performance).

Is this the correct performance?

the command I used for these experiments is (normally i just modify the
rate and the weight of the class):

tc qdisc add dev eth0 root handle 1: cbq bandwidth 10Mbit cell 8 avpkt \
1000 mpu 64

tc class add dev eth0 parent 1:0 classid 1:1 cbq bandwidth 10Mbit rate \
10Mbit allot 1514 cell 8 weight 1Mbit prio 8 maxburst 20 avpkt 1000

tc class add dev eth0 parent 1:1 classid 1:2 cbq bandwidth 10Mbit rate \
100Kbit bounded allot 1514 cell 8 weight 10Kbit prio 5 maxburst 20 avpkt \
1000 split 1:0

tc filter add dev eth0 parent 1:0 protocol ip prio 100 route to 1 \
classid 1:2

ip route add 161.142.8.21 dev eth0 via 10.39.11.23 realm 1

Thanks for any help or comment! I appreciate! Really!

Regards,
Chin Hooi

-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to majordomo@vger.rutgers.edu