Hello, I created an application using ONOS that creates the queues via OVSDB and sets the flowrules to redirect the selected traffic to those queues, the problem now is that the delay is similar for any packet that passes, which means either a packet have delay x or dropped, which makes me thinking, is it possible that the packets are not queued ?
Here is some data using traffic control : class htb 1:fffe root rate 1Mbit ceil 1Mbit linklayer ethernet burst 1500b/1 mpu 0b overhead 0b cburst 1500b/1 mpu 0b overhead 0b level 7 Sent 6746523 bytes 122412 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 lended: 46985 borrowed: 0 giants: 0 tokens: 154480 ctokens: 154480 class htb 1:1 parent 1:fffe prio 7 quantum 1500 rate 12Kbit ceil 1Mbit linklayer ethernet burst 1563b/1 mpu 0b overhead 0b cburst 1564b/1 mpu 0b overhead 0b level 0 Sent 223323 bytes 1612 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 lended: 1612 borrowed: 0 giants: 0 tokens: 13397562 ctokens: 162480 class htb 1:2 parent 1:fffe prio 1 quantum 10000 rate 800Kbit ceil 1Mbit linklayer ethernet burst 4000b/1 mpu 0b overhead 0b cburst 4000b/1 mpu 0b overhead 0b level 0 Sent 2700000 bytes 50000 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 lended: 50000 borrowed: 0 giants: 0 tokens: 616562 ctokens: 493250 class htb 1:3 parent 1:fffe prio 2 quantum 1875 rate 150Kbit ceil 1Mbit linklayer ethernet burst 1563b/1 mpu 0b overhead 0b cburst 1564b/1 mpu 0b overhead 0b level 0 Sent 2700000 bytes 50000 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 lended: 17875 borrowed: 32125 giants: 0 tokens: -8702 ctokens: 188750 class htb 1:4 parent 1:fffe prio 3 quantum 1500 rate 49Kbit ceil 1Mbit linklayer ethernet burst 1563b/1 mpu 0b overhead 0b cburst 1564b/1 mpu 0b overhead 0b level 0 Sent 1123200 bytes 20800 pkt (dropped 29200, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 lended: 5940 borrowed: 14860 giants: 0 tokens: -41293 ctokens: 188746 as you can see htb 1:2 should have the highest priority, and takes at least 80% of the bandwidth. the problem is that when i send traffic with different priorities to be directed to the three queues (1:2 1:3 1:4) all the packets that passes the queues have almost the same delay and queuing time ! it shouldn't be like that, my vision is to make the highest priority queue send first and the other packets stays on the queues and be sent after the queue 1:2 finishes sending, and with that the latency will be different, but somehow this does not happens so far. Three pings were sent simultaneous with different priorities (so they can go through the selected classes), i sent 50.000 packets with a rate of 1000 packet/s for each queue (150.000 packet to the the class 1) and here are the results : htb 1:2 Max rtt: 3.725ms | Min rtt: 0.010ms | Avg rtt: 0.010ms Raw packets sent: 50000 (2.000MB) | Rcvd: 41086 (1.643MB) | Lost: 8914 (17.83%) htb 1:3 Max rtt: 1.745ms | Min rtt: 0.009ms | Avg rtt: 0.009ms Raw packets sent: 50000 (2.000MB) | Rcvd: 41904 (1.676MB) | Lost: 8096 (16.19%) htb 1:4 Max rtt: 4.539ms | Min rtt: 0.009ms | Avg rtt: 0.010ms Raw packets sent: 50000 (2.000MB) | Rcvd: 16605 (664.200KB) | Lost: 33395 (66.79%) the queues on the switch : ubuntu@tsn-server:~$ sudo ovs-ofctl queue-stats s1 OFPST_QUEUE reply (xid=0x2): 8 queues port "s1-eth1" queue 0: bytes=231663, pkts=1672, errors=0, duration=? port "s1-eth1" queue 1: bytes=2700000, pkts=50000, errors=0, duration=? port "s1-eth1" queue 2: bytes=2700000, pkts=50000, errors=0, duration=? port "s1-eth1" queue 3: bytes=1123200, pkts=20800, errors=0, duration=? port "s1-eth2" queue 0: bytes=231941, pkts=1674, errors=0, duration=? port "s1-eth2" queue 1: bytes=2700000, pkts=50000, errors=0, duration=? port "s1-eth2" queue 2: bytes=2700000, pkts=50000, errors=0, duration=? port "s1-eth2" queue 3: bytes=1123200, pkts=20800, errors=29200, duration=? please help me out ! Thank you.
_______________________________________________ discuss mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-discuss
