Hi, all
I am interested in Linux related QoS, especially PRIO queue discipline (qdisc).
I test some experiments about PRIO qdisc.
Host A ------ Router A --- Router B --- Host C
| 1Gbps |
Host B ---------+ +-------- Host D
flow 1 : Host A -> Host C (TCP)
flow 2 : Host B -> Host D (UDP)
priority 1 queue (pfifo, size 1000 packets ) : flow 1
priority 2 queue (pfifo, size 1000 packets ) : flow 2
priority 3 queue (pfifo, size 1000 packets ) : defaults
I configured PRIO qdisc on Router A's outer interface to Router B.
the results of test
- TCP throughput about 80Mbps
- UDP throughput about 900Mbps. (UDP try to send 1Gbps)
First question:
The TCP stream with higher priority than UDP stream with lower priority
experienced starvation in stead of UDP stream. Is it correct?
Did you test PRIO qdisc with TCP having high priority and UDP having low priority?
In the below, there is my configuration. Is there something wrong?
Second Question:
When I test with one TCP stream having high priority and one TCP stream having
low priority, the throughput of each TCP stream is about 500Mbps.
When I increase the number of TCP stream, the TCP streams having high priority
have more bandwidth than TCP streams with low priority. (see below,B.TCP vs TCP)
Is it correct?
Thanks in advance,
Sincerely,
Sangho Lee
----------------------------------------------------------------------------------
A. My Configuration
tc qdisc add dev eth1 root handle 1: prio bands 3 priomap 2 2 2 2 2 2 2 2 2 2 2
2 2 2 2 2
tc qdisc add dev eth1 parent 1:1 handle 10: pfifo
tc qdisc add dev eth1 parent 1:2 handle 20: pfifo
tc qdisc add dev eth1 parent 1:3 handle 30: pfifo
tc filter add dev eth1 parent 1: protocol ip prio 1 u32 match ip src
134.77.1.151/32 match ip dst 134.75.1.151/32 flowid 1:1
tc filter add dev eth1 parent 1: protocol ip prio 2 u32 match ip src
134.78.1.151/32 match ip dst 134.76.1.151/32 flowid 1:2
B. TCP vs TCP
1) 1 stream : 462 vs 415 Mbps
2) 3 streams : 465 vs 482 Mbps
3) 10 streams : 477 vs 461 Mbps
4) 30 streams : 76 vs 864 Mbps
5) 100 streams : 35.3 vs 911 Mbps
_______________________________________________
LARTC mailing list
LARTC@xxxxxxxxxxxxxxx
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
