[ovs-discuss] High CPU Usage by ovs-vswitchd

[Jose Alberto Posada]

Hello I need help,
We have a setup in which a core network server running. We have
installed openvswitch (1.9.0) in the same server (running Ubuntu-13.04
raring server 64bit).
The CPU usage shoots up to 98-100% (on a quad core Intel(R) Xeon(R)
CPU E5420  @ 2.50GHz), here is the configuration:

~$ sudo ovs-vsctl show
758c9ac6-c35f-48ff-be24-b1722d963c81
    Bridge "br0"
        Port "bond12"
            trunks: [7, 9, 59, 102, 113, 208, 301, 302, 303, 304, 305,
306, 307, 308, 309, 310, 651, 652, 700, 701]
            Interface "eth2"
            Interface "eth1"
        Port "br0"
            tag: 701
            Interface "br0"
                type: internal
        Port "veth12"
            trunks: [7, 9, 59, 102, 113, 208, 301, 302, 303, 304, 305,
306, 307, 308, 309, 310, 651, 652, 700, 701]
            Interface "veth12"
        Port "veth34"
            trunks: [5, 6, 8, 11, 49, 50, 53, 54, 56, 58, 89, 90, 144,
181, 182, 183, 184, 190, 191, 192, 193, 194, 195, 244, 245, 702, 1080]
            Interface "veth34"
        Port "eth0"
            trunks: [81, 82, 83, 104, 106, 110, 128, 129, 131, 132,
133, 141, 142, 143, 160, 176, 242, 243, 650]
            Interface "eth0"
        Port "bond34"
            trunks: [5, 6, 8, 11, 49, 50, 53, 54, 56, 58, 89, 90, 144,
181, 182, 183, 184, 190, 191, 192, 193, 194, 195, 244, 245, 702, 1080]
            Interface "eth4"
            Interface "eth3"
        Port "veth0"
            trunks: [81, 82, 83, 104, 106, 110, 128, 129, 131, 132,
133, 141, 142, 143, 160, 176, 242, 243, 650]
            Interface "veth0"
    ovs_version: "1.9.0"

~$ sudo ovs-dpctl show
system at br0:
        lookups: hit:367280347 missed:17177097 lost:517505
        flows: 1145
        port 0: br0 (internal)
        port 1: eth1
        port 2: eth2
        port 3: eth3
        port 4: eth0
        port 5: eth4
        port 12: veth0
        port 13: veth12
        port 14: veth34

ovs-vsctl set bridge br0 other_config:flow-eviction-threshold=10000

Is there anything else we can do to reduce CPU usage?
Is there any other configuration for deployment in production that we
missed?

Best regards
Jose Posada