Problem configuring QoS with FlowVisor and OpenVswitch

Question

I am running a small experiment to isolate bandwidth with flowvisor (version 1.4, version 1.0 of openflow) implementing a port based slicing. With mininet, I create a topology that consists of a host h1 (10.0.0.1), connected to a switch, called s3, connected to another host, h2 (10.0.0.2), this way:

h1 ----- s3 ------ h2

To achieve this I use this mininet command to run my custom topology (called simpletopo.py): sudo mn --custom simpletopo.py --topo mytopo --switch ovsk --controller remote --mac --arp --link tc

I want all traffic going to port 666 from h1 to h2 (and vice versa) to have a bandwidth of 1Mbps, and all traffic going to port 555 from h1 to h2 (and vice versa) to have a bandwidth of 1Gbps. To achieve this, I create two queues in switch s3, q0 and q1. q0 is for the 1Mbps traffic and q1 for 1Gbps traffic.

First, I create the two slices:

fvctl add-slice slice1 tcp:127.0.0.2:6643 [email protected]
fvctl add-slice slice2 tcp:127.0.0.3:6653 [email protected]

Then, I allocate flowspace:

fvctl -f pass add-flowspace --queues=1 --forced-queue=1 flow1 3 100 in_port=2,tp_dst=666 slice1=7
fvctl -f pass add-flowspace --queues=1 --forced-queue=1 flow2 3 100 in_port=1,tp_src=666 slice1=7
fvctl -f pass add-flowspace --queues=1 --forced-queue=1 flow3 3 100 in_port=1,tp_dst=666 slice1=7
fvctl -f pass add-flowspace --queues=1 --forced-queue=1 flow4 3 100 in_port=2,tp_src=666 slice1=7
fvctl -f pass add-flowspace --queues=0 --forced-queue=0 flow5 3 100 in_port=2,tp_dst=555 slice2=7
fvctl -f pass add-flowspace --queues=0 --forced-queue=0 flow6 3 100 in_port=1,tp_src=555 slice2=7
fvctl -f pass add-flowspace --queues=0 --forced-queue=0 flow7 3 100 in_port=1,tp_dst=555 slice2=7
fvctl -f pass add-flowspace --queues=0 --forced-queue=0 flow8 3 100 in_port=2,tp_src=555 slice2=7

I use the queue option to tell flowvisor which slices are available to each flowspace and the forced-queue option to force flowvisor to change the output actions for enqueue actions in the queue specified.

Then I run two POX controllers, one for each slice:

./pox.py --verbose pox.openflow.of_01 --address=127.0.0.2 --port=6643 pox.forwarding.l2_learning
./pox.py --verbose pox.openflow.of_01 --address=127.0.0.3 --port=6653 pox.forwarding.l2_learning

Now, I run the topology with mininet. At this point, the two slices work fine, each one controlls its traffic. Traffic uses all bandwidth.

Now I want to throttle the bandwidth. To achieve this I create the queues mentioned above.

I create the queues like this:

sudo ovs-vsctl -- set Port s3-eth1 qos=@newqos -- set Port s3-eth2 qos=@newqos -- --id=@newqos create QoS type=linux-htb other-config:max-rate=1000000000 queues=0=@q0,1=@q1 -- --id=@q0 create Queue other-config:min-rate=1000000 other-config:max-rate=1000000 -- --id=@q1 create Queue other-config:min-rate=1000000000 other-config:max-rate=1000000000

When I run tests with iperf, I got all traffic throttled to 1 Mbps. When I run sudo ovs-ofctl queue-stats s3, I see that all traffic passes through queue 0 (default queue for openflow) and no traffic passes through queue 1.

What's going wrong? Is there something wrong with flowvisor or the creation of queues?

Thanks a lot.

Answer 1

Finally I figured out what was wrong:

After creating the qos and queues, it's necessary to define the flows, like this:

sudo ovs-ofctl add-flow s3 in_port=1,dl_type=0x0800,nw_proto=6,tp_dst=666,priority=100,actions=set_queue:1,normal

sudo ovs-ofctl add-flow s3 in_port=1,dl_type=0x0800,nw_proto=6,tp_dst=555,priority=100,actions=set_queue:0,normal

sudo ovs-ofctl add-flow s3 in_port=2,dl_type=0x0800,nw_proto=6,tp_dst=666,priority=100,actions=set_queue:1,normal

sudo ovs-ofctl add-flow s3 in_port=2,dl_type=0x0800,nw_proto=6,tp_dst=555,priority=100,actions=set_queue:0,normal

This way we are telling openflow which traffic must pass through each queue. However, with this procedure, I have seen that it is not necessary to use the options queue and forced-queue of flowvisor.

Source:

ovs-ofctl add-flow<bridge><match-field>actions=set_queue:<queue>

By taking this action, the packets that is output the port will be output the specified queue. Different switch has different numbers of supported queues. (http://www.pica8.com/document/v2.3/html/ovs-commands-reference/)