[ovs-dev,net-next,v3,00/10] optimize openvswitch flow looking up
mbox series

Message ID 1570802447-8019-1-git-send-email-xiangxia.m.yue@gmail.com
Headers show
Series
  • optimize openvswitch flow looking up
Related show

Message

Tonghao Zhang Oct. 11, 2019, 2 p.m. UTC
From: Tonghao Zhang <xiangxia.m.yue@gmail.com>

This series patch optimize openvswitch for performance or simplify
codes.

Patch 1, 2, 4: Port Pravin B Shelar patches to
linux upstream with little changes.

Patch 5, 6, 7: Optimize the flow looking up and
simplify the flow hash.

Patch 8, 9: are bugfix.

The performance test is on Intel Xeon E5-2630 v4.
The test topology is show as below:

+-----------------------------------+
|   +---------------------------+   |
|   | eth0   ovs-switch    eth1 |   | Host0
|   +---------------------------+   |
+-----------------------------------+
      ^                       |
      |                       |
      |                       |
      |                       |
      |                       v
+-----+----+             +----+-----+
| netperf  | Host1       | netserver| Host2
+----------+             +----------+

We use netperf send the 64B packets, and insert 255+ flow-mask:
$ ovs-dpctl add-flow ovs-switch "in_port(1),eth(dst=00:01:00:00:00:00/ff:ff:ff:ff:ff:01),eth_type(0x0800),ipv4(frag=no)" 2
...
$ ovs-dpctl add-flow ovs-switch "in_port(1),eth(dst=00:ff:00:00:00:00/ff:ff:ff:ff:ff:ff),eth_type(0x0800),ipv4(frag=no)" 2
$
$ netperf -t UDP_STREAM -H 2.2.2.200 -l 40 -- -m 18

* Without series patch, throughput 8.28Mbps
* With series patch, throughput 46.05Mbps

v2: simplify codes. e.g. use kfree_rcu instead of call_rcu, use
ma->count in the fastpath.
v3: update ma point when realloc mask_array in patch 5.

Tonghao Zhang (10):
  net: openvswitch: add flow-mask cache for performance
  net: openvswitch: convert mask list in mask array
  net: openvswitch: shrink the mask array if necessary
  net: openvswitch: optimize flow mask cache hash collision
  net: openvswitch: optimize flow-mask looking up
  net: openvswitch: simplify the flow_hash
  net: openvswitch: add likely in flow_lookup
  net: openvswitch: fix possible memleak on destroy flow-table
  net: openvswitch: don't unlock mutex when changing the user_features
    fails
  net: openvswitch: simplify the ovs_dp_cmd_new

 net/openvswitch/datapath.c   |  65 +++++----
 net/openvswitch/flow.h       |   1 -
 net/openvswitch/flow_table.c | 315 +++++++++++++++++++++++++++++++++++++------
 net/openvswitch/flow_table.h |  19 ++-
 4 files changed, 328 insertions(+), 72 deletions(-)

Comments

Greg Rose Oct. 15, 2019, 8:01 p.m. UTC | #1
On 10/11/2019 7:00 AM, xiangxia.m.yue@gmail.com wrote:
> From: Tonghao Zhang <xiangxia.m.yue@gmail.com>
>
> This series patch optimize openvswitch for performance or simplify
> codes.
>
> Patch 1, 2, 4: Port Pravin B Shelar patches to
> linux upstream with little changes.
>
> Patch 5, 6, 7: Optimize the flow looking up and
> simplify the flow hash.
>
> Patch 8, 9: are bugfix.
>
> The performance test is on Intel Xeon E5-2630 v4.
> The test topology is show as below:
>
> +-----------------------------------+
> |   +---------------------------+   |
> |   | eth0   ovs-switch    eth1 |   | Host0
> |   +---------------------------+   |
> +-----------------------------------+
>        ^                       |
>        |                       |
>        |                       |
>        |                       |
>        |                       v
> +-----+----+             +----+-----+
> | netperf  | Host1       | netserver| Host2
> +----------+             +----------+
>
> We use netperf send the 64B packets, and insert 255+ flow-mask:
> $ ovs-dpctl add-flow ovs-switch "in_port(1),eth(dst=00:01:00:00:00:00/ff:ff:ff:ff:ff:01),eth_type(0x0800),ipv4(frag=no)" 2
> ...
> $ ovs-dpctl add-flow ovs-switch "in_port(1),eth(dst=00:ff:00:00:00:00/ff:ff:ff:ff:ff:ff),eth_type(0x0800),ipv4(frag=no)" 2
> $
> $ netperf -t UDP_STREAM -H 2.2.2.200 -l 40 -- -m 18
>
> * Without series patch, throughput 8.28Mbps
> * With series patch, throughput 46.05Mbps
>
> v2: simplify codes. e.g. use kfree_rcu instead of call_rcu, use
> ma->count in the fastpath.
> v3: update ma point when realloc mask_array in patch 5.
>
> Tonghao Zhang (10):
>    net: openvswitch: add flow-mask cache for performance
>    net: openvswitch: convert mask list in mask array
>    net: openvswitch: shrink the mask array if necessary
>    net: openvswitch: optimize flow mask cache hash collision
>    net: openvswitch: optimize flow-mask looking up
>    net: openvswitch: simplify the flow_hash
>    net: openvswitch: add likely in flow_lookup
>    net: openvswitch: fix possible memleak on destroy flow-table
>    net: openvswitch: don't unlock mutex when changing the user_features
>      fails
>    net: openvswitch: simplify the ovs_dp_cmd_new
>
>   net/openvswitch/datapath.c   |  65 +++++----
>   net/openvswitch/flow.h       |   1 -
>   net/openvswitch/flow_table.c | 315 +++++++++++++++++++++++++++++++++++++------
>   net/openvswitch/flow_table.h |  19 ++-
>   4 files changed, 328 insertions(+), 72 deletions(-)
>

Hi Tonghao,

I've tried this new patch series and it passes the kernel check test #63 
now:
## ------------------------------- ##
## openvswitch 2.12.90 test suite. ##
## ------------------------------- ##
  63: conntrack - IPv6 fragmentation + vlan           ok

## ------------- ##
## Test results. ##
## ------------- ##

1 test was successful.

So I went ahead and ran the entire check-kernel testsuite and it ran 
fine with no regressions or
other problems.

You can go ahead and add my tested by tag to your patches.
Tested-by: Greg Rose <gvrose8192@gmail.com>

Pravin's comments about the memory barrier are still valid I think.

Thanks,

- Greg