On 1/5/23 21:24, David Marchand wrote:
> The DPDK vhost-user library maintains more granular per queue stats
> which can replace what OVS was providing for vhost-user ports.
>
> The benefits for OVS:
> - OVS can skip parsing packet sizes on the rx side,
> - dev->stats_lock won't be taken in rx/tx code unless some packet is
> dropped,
> - vhost-user is aware of which packets are transmitted to the guest,
> so per *transmitted* packet size stats can be reported,
> - more internal stats from vhost-user may be exposed, without OVS
> needing to understand them,
>
> Note: the vhost-user library does not provide global stats for a port.
> The proposed implementation is to have the global stats (exposed via
> netdev_get_stats()) computed by querying and aggregating all per queue
> stats.
> Since per queue stats are exposed via another netdev ops
> (netdev_get_custom_stats()), this may lead to some race and small
> discrepancies.
> This issue might already affect other netdev classes.
>
> Example:
> $ ovs-vsctl get interface vhost4 statistics |
> sed -e 's#[{}]##g' -e 's#, #\n#g' |
> grep -v =0$
> rx_1_to_64_packets=12
> rx_256_to_511_packets=15
> rx_65_to_127_packets=21
> rx_broadcast_packets=15
> rx_bytes=7497
> rx_multicast_packets=33
> rx_packets=48
> rx_q0_good_bytes=242
> rx_q0_good_packets=3
> rx_q0_guest_notifications=3
> rx_q0_multicast_packets=3
> rx_q0_size_65_127_packets=2
> rx_q0_undersize_packets=1
> rx_q1_broadcast_packets=15
> rx_q1_good_bytes=7255
> rx_q1_good_packets=45
> rx_q1_guest_notifications=45
> rx_q1_multicast_packets=30
> rx_q1_size_256_511_packets=15
> rx_q1_size_65_127_packets=19
> rx_q1_undersize_packets=11
> tx_1_to_64_packets=36
> tx_256_to_511_packets=45
> tx_65_to_127_packets=63
> tx_broadcast_packets=45
> tx_bytes=22491
> tx_multicast_packets=99
> tx_packets=144
> tx_q0_broadcast_packets=30
> tx_q0_good_bytes=14994
> tx_q0_good_packets=96
> tx_q0_guest_notifications=96
> tx_q0_multicast_packets=66
> tx_q0_size_256_511_packets=30
> tx_q0_size_65_127_packets=42
> tx_q0_undersize_packets=24
> tx_q1_broadcast_packets=15
> tx_q1_good_bytes=7497
> tx_q1_good_packets=48
> tx_q1_guest_notifications=48
> tx_q1_multicast_packets=33
> tx_q1_size_256_511_packets=15
> tx_q1_size_65_127_packets=21
> tx_q1_undersize_packets=12
>
> Reviewed-by: Maxime Coquelin <[email protected]>
> Signed-off-by: David Marchand <[email protected]>
> ---
> Changes since v5:
> - added missing dev->stats_lock acquire in netdev_dpdk_vhost_get_stats,
> - changed netdev_dpdk_vhost_update_[rt]x_counters to take dev->stats_lock
> only when some packets got dropped in OVS. Since the rx side won't
> take the lock unless some QoS configuration is in place, this change
> will likely have the same effect as separating stats_lock into rx/tx
> dedicated locks. Testing shows a slight (around 1%) improvement of
> performance for some V2V setup,
Thanks, David. With this version I see a performance improvement
over the master branch as well. Nice!
Maxime, could you take another look at this series? Or did you
already? I see your 'Reviewed-by' tag, but I'm not sure if it's
just a carry over.
Best regards, Ilya Maximets.
>
> Changes since v3:
> - rebased to master now that v22.11 landed,
> - fixed error code in stats helper when vhost port is not "running",
> - shortened rx/tx stats macro names,
>
> Changes since RFC v2:
> - dropped the experimental api check (now that the feature is marked
> stable in DPDK),
> - moved netdev_dpdk_get_carrier() forward declaration next to the
> function needing it,
> - used per q stats for netdev_get_stats() and removed OVS per packet
> size accounting logic,
> - fixed small packets counter (see rx_undersized_errors hack),
> - added more Tx stats,
> - added unit tests,
>
> ---
> lib/netdev-dpdk.c | 429 ++++++++++++++++++++++++++++++++-----------
> tests/system-dpdk.at | 33 +++-
> 2 files changed, 349 insertions(+), 113 deletions(-)
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
