TCP UPD performance test

Report TCP or UDP performance

Get the performance statistics of TCP UDP traffic and report it.

1. For each iteration (i = 0; i < n_iut_ports; ++i):

   • Configure Rx checksum offload on IUT interface if specified

   • Configure GRO on IUT interface if specified

   • Configure HW VLAN stripping on IUT interface if specified

   • Configure Tx checksum offload on IUT interface if specified

   • Configure GSO offload on IUT interface if specified

   • Configure TSO offload on IUT interface if specified

   • Configure HW VLAN insertion on IUT interface if specified

   • If rx_coalesce_usecs or rx_max_coalesced_frames is not -1, configure Rx coalesce on IUT interface.

     • If rx_coalesce_usecs != -1 || rx_max_coalesced_frames != -1 :

       • Disable use_adaptive_rx_coalesce on IUT interface.

       • If rx_coalesce_usecs != -1 :

         • If rx_coalesce_usecs is not -1, configure it IUT interface.

       • If rx_max_coalesced_frames != -1 :

         • If rx_max_coalesced_frames is not -1, configure it IUT interface.

   • If rx_ring != -1 :

     • Set Rx ring size according to rx_ring on IUT interface.

   • If tx_ring != -1 :

     • Set Tx ring size according to tx_ring on IUT interface.

   • If channels != -1 :

     • Set number of combined channels on IUT interface according to channels.

       • Get current number of Rx queues on IUT interface

       • If current number of Rx queues is more than number channels to be set, update RSS indirection table to use only Rx queues which remain after channels set.

       • Apply number of combined channels on IUT interface according to channels.

       • If previously used number of Rx queues is less than number of channels, update RSS indirection table to use all Rx queues available after channels set.

       • Print RSS indirection table to be able to double-check.

2. For each iteration (i = 0; i < n_ports; i++):

   • If rx_vlan_strip or tx_vlan_insert is not default, create VLANs, assign addresses and use it for traffic checks below.

3. Set default perf options

4. Set test specific perf options

5. Allocate server ports for perf applications

6. Start server and create client perf applications

7. Start perf clients

8. Wait for perf report to be ready

Igor Romanov Igor.Romanov@oktetlabs.ru

Parameters:

rx_csum	Enable (TRUE), disable (FALSE) Rx checksum offload or preserve default (DEFAULT) Possible values: DEFAULT, FALSE, TRUE
rx_gro	Enable (TRUE), disable (FALSE) Rx GRO offload or preserve default (DEFAULT) Possible values: DEFAULT, FALSE, TRUE
rx_vlan_strip	Enable (TRUE), disable (FALSE) Rx VLAN stripping offload or preserve default (DEFAULT) Possible values: DEFAULT, FALSE, TRUE
tx_csum	Enable (TRUE), disable (FALSE) Tx checksum offload or preserve default (DEFAULT) Possible values: DEFAULT, FALSE, TRUE
tx_gso	Enable (TRUE), disable (FALSE) Tx GSO offload or preserve default (DEFAULT) Possible values: DEFAULT, FALSE, TRUE
tso	Enable (TRUE), disable (FALSE) TSO offload or preserve default (DEFAULT) Possible values: DEFAULT, FALSE, TRUE
tx_vlan_insert	Enable (TRUE), disable (FALSE) Tx VLAN insertion offload or preserve default (DEFAULT) Possible values: DEFAULT, FALSE, TRUE
rx_coalesce_usecs	Value to set for rx_coalesce_usecs: -1 : keep default settings 0 : interrupt moderation is disabled 30 150
rx_max_coalesced_frames	Value to set for rx_max_coalesced_frames: -1 : keep default settings 0 : do not coalesce based on it 1 : interrupt moderation is disabled
rx_ring	Rx rings size: -1 : keep default 0 : maximum 64 512 1024
tx_ring	Tx rings size: -1 : keep default 0 : maximum 64 512 1024
channels	Number of combined channels to use: -1 : keep default 1 2 4
perf_bench	Performance benchmark type
dual_mode	Run benchmark in bidirectional mode
n_perf_insts	Number of performance apps to run
n_streams	Number of parallel streams to run
bandwidth	Target bandwidth in Mbps, negative means some internal tool value for UDP, unlimited for TCP
protocol	Use TCP or UDP protocol