[Enterprise SONiC] QoS (Quality of Service) Follow
Restriction:
-
- The default trust mode of the port is dot1p.
If the port is configured DSCP-to-TC mapping, the trust mode of the port will change to DSCP. - QoS Marking function only supports DCS503 (AS9716-32D)
- QoS scheduling and shaping is only supported on physical interfaces, physical interface egress queues, and unicast queues, it's not supported on VLAN and port-channel interfaces.
- WRR scheduling is not supported on Intel switches.
- QoS scheduling default is WRR on the Broadcom platform, the weight of each queue is 1.
- There is no minimum bandwidth guarantee support for queue shaping on Broadcom and Intel switches.
- There are 8 unicast egress queues (UC0 to UC7) supported; multicast queues are not supported. Some platforms may display UC8 and UC9, but neither egress queue is supported
- The default trust mode of the port is dot1p.
QoS Classification
QoS Classification is used to classify the packets to assign a Traffic Class (TC) value based on service requirements. TC is the internal priority in switch (8 possible value) using for packet buffering and scheduling. TC-to-priority-group (PG) and TC-to-Queue maps depend on the TC maps to the ingress buffer (priority-group) and egress queue, respectively. TC is used as a priority field for packet re-marking on egress port. Determines the TC based on the priority of the incoming packet and trust mode of a port.
-
Dot1p-to-TC Mapping is assigned the TC based on PCP（Priority Code Point）in the VLAN tag, and this mapping table is configured per ingress physical port.
Default Dot1p-to-TC mapping on Broadcom platform:Dot1P 0 1 2 3 4 5 6 7 TC 0 1 2 3 4 5 6 7 Default Dot1p-to-TC mapping on the Intel platform:
Dot1P 0~7 TC 0 -
DSCP-to-TC Mapping is assigned the TC based on DSCP field in the IP header, and this mapping table is configured per ingress physical port.
Default DSCP-to-TC mapping:DSCP 0~63 TC 0 -
TC-to-Queue Mapping is assigned the Queue ID based on TC, and this mapping table is configured per ingress physical port.
Default TC-to-Queue mapping on Broadcom platform:TC 0 1 2 3 4 5 6 7 Queue 0 1 2 3 4 5 6 7 Default TC-to-Queue mapping on Intel platform:
TC 0~7 Queue 0
Tested model & firmware version:
- Switch model name:
DCS503 (AS9716-32D)
- Edgecore SONiC version:
202111.1
202111.3
202111.8
Topology:
Pre-configuration:
Add VLAN 10 to Ethernet0, Ethernet8.
admin@sonic:~$ show vlan brief
+-----------+--------------+-----------+----------------+-------------+---------------+---------------+-------------+
| VLAN ID | IP Address | Ports | Port Tagging | Proxy ARP | DHCP Helper | DHCP Source | DHCP Link |
| | | | | | Address | Interface | Selection |
+===========+==============+===========+================+=============+===============+===============+=============+
| 10 | | Ethernet0 | tagged | disabled | | | |
| | | Ethernet8 | tagged | | | | |
+-----------+--------------+-----------+----------------+-------------+---------------+---------------+-------------+
Procedure:
Step 1: Create a profile for DOT1P/DSCP to TC(Traffic Class).
Example for Dot1P:
admin@sonic:~$ sudo config qos dot1p-tc add 1p_tc --dot1p 0 --tc 1
Example for DCSP:
admin@sonic:~$ sudo config qos dscp-tc add DSCP_TC --dscp 7 --tc 1
Note:
admin@sonic:~$ sudo config qos dot1p-tc add --help
Usage: config qos dot1p-tc add [OPTIONS] <profile>
Add dot1p-tc map profile.
Options:
--dot1p TEXT Cos value [required]
--tc INTEGER RANGE Traffic-class(TC) value [required]
-h, -?, --help Show this message and exit.
Step 2: Modify the existing Dot1p/DSCP to TC profile.
Example for DOT1P:
admin@sonic:~$ sudo config qos dot1p-tc update 1p_tc --dot1p 1 --tc 2
Example for DCSP:
admin@sonic:~$ sudo config qos dscp-tc update DSCP_TC --dscp 8 --tc 2
Note:
admin@sonic:~$ sudo config qos dot1p-tc update --help
Usage: config qos dot1p-tc update [OPTIONS] <profile>
Update dot1p-tc map profile.
Options:
--dot1p TEXT Cos value [required]
--tc INTEGER RANGE Traffic-class(TC) value
--remove Delete the mapping entry
-?, -h, --help Show this message and exit.
Step 3: Check the profile of DOT1P/DSCP to TC.
DOT1P to TC:
admin@sonic:~$ show qos dot1p-tc
dot1p-tc policy: 1p_tc
Dot1p TC
------- ----
0 1
1 2
DSCP to TC:
admin@sonic:~$ show qos dscp-tc
dscp-tc policy: DSCP_TC
DSCP TC
------ ----
7 1
8 2
Step 4: Create a profile for TC to Queue.
admin@sonic:~$ sudo config qos tc-queue add TC_Q --tc 1 --queue 2
Note:
admin@sonic:~$ sudo config qos tc-queue add --help
Usage: config qos tc-queue add [OPTIONS] <profile>
Add tc-queue map profile.
Options:
--tc TEXT Traffic-class(TC) value [required]
--queue INTEGER RANGE Queue value [required]
-?, -h, --help Show this message and exit.
Step 5: Modify the existing TC to Queue profile.
admin@sonic:~$ sudo config qos tc-queue update TC_Q --tc 2 --queue 3
Note:
admin@sonic:~$ sudo config qos tc-queue update --help
Usage: config qos tc-queue update [OPTIONS] <profile>
Update tc-queue map profile.
Options:
--tc TEXT Traffic-class(TC) value [required]
--queue INTEGER RANGE Queue value
--remove Delete the mapping entry
-?, -h, --help Show this message and exit.
Step 6: Check the profile of TC to Queue.
admin@sonic:~$ show qos tc-queue
tc-queue policy: TC_Q
TC Queue
---- -------
1 2
2 3
Step 7: Binding the mapping table to the specified interface.
admin@sonic:~$ sudo config interface qos dot1p-tc bind Ethernet0 1p_tc
admin@sonic:~$ sudo config interface qos dscp-tc bind Ethernet0 DSCP_TC
admin@sonic:~$ sudo config interface qos tc-queue bind Ethernet0 TC_Q
Note:
admin@sonic:~$ sudo config interface qos dot1p-tc --help
Usage: config interface qos dot1p-tc [OPTIONS] <op> <interface_name> <profile>
dot1p-tc policy configuration
Options:
-?, -h, --help Show this message and exit.
- "op": "bind/unbind"
Step 6: Check the binding table.
admin@sonic:~$ show interfaces qos
Ethernet0:
Dot1p to TC: 1p_tc
DSCP to TC: DSCP_TC
TC to Queue: TC_Q
Result: Check the queue counters
When the VLAN 10 packet with priority 0 into the switch, it will follow the mapping table that we set.
Dot1p ------ > TC ------ > Queue
Based on the above setting, the VLAN 10 packet with priority 0 will be forwarded to Queue2.
Step 1. Clear the queue counter
Step 2. Check Ethernet8 (egress port) queue counter
QoS Marking
Once the packets have been classified by QoS Classification, QoS Marking allows to set or change the priority value in the packets before sending them. The new priority value is brought to the next switches for fine-tuning or uniform the class of the traffic in the network.
QoS Marking enables to rewrite the Dot1p or DSCP of incoming packets based on Traffic Class (TC) value.
-
Enable Dot1p rewrite by configuring a TC-to-Dot1p mapping on an egress port. On Broadcom switches, the Dot1p rewrite is always enabled.
Default TC-to-Dot1p mapping.TC 0 1 2 3 4 5 6 7 Dot1P 0 1 2 3 4 5 6 7 - Enable DSCP rewrite by configuring a TC-to-DSCP mapping on an egress port. The DSCP rewrite is disabled by default on all ports.
Tested model & firmware version:
- Switch model name:
DCS503 (AS9716-32D)
- Edgecore SONiC version:
202111.1
202111.3
Topology:
Pre-configuration:
Add VLAN 10 to Ethernet0, Ethernet8.
admin@sonic:~$ show vlan brief
+-----------+--------------+-----------+----------------+-------------+---------------+---------------+-------------+
| VLAN ID | IP Address | Ports | Port Tagging | Proxy ARP | DHCP Helper | DHCP Source | DHCP Link |
| | | | | | Address | Interface | Selection |
+===========+==============+===========+================+=============+===============+===============+=============+
| 10 | | Ethernet0 | tagged | disabled | | | |
| | | Ethernet8 | tagged | | | | |
+-----------+--------------+-----------+----------------+-------------+---------------+---------------+-------------+
Procedure:
Step 1: Create a profile for DOT1P remarking.
Example for Dot1P:
admin@sonic:~$ sudo config qos remark dot1p add remark_dot1p --tc 0 --dot1p 1
Example for DSCP:
admin@sonic:~$ sudo config qos remark dscp add remark_dscp --tc 0 --dscp 7
Note:
admin@sonic:~$ sudo config qos remark dot1p add --help
Usage: config qos remark dot1p add [OPTIONS] <profile>
Add tc-dot1p map profile.
Options:
--tc TEXT Traffic-class(TC) value [required]
--dot1p INTEGER RANGE Cos value [required]
-?, -h, --help Show this message and exit.
Step 2: Modify the existing remark profile.
Example for Dot1P:
admin@sonic:~$ sudo config qos remark dot1p update remark_dot1p --tc 1 --dot1p 2
Example for DSCP:
admin@sonic:~$ sudo config qos remark dscp update remark_dscp --tc 1 --dscp 8
Note:
admin@sonic:~$ sudo config qos remark dot1p update --help
Usage: config qos remark dot1p update [OPTIONS] <profile>
Update tc-dot1p map profile.
Options:
--tc TEXT Traffic-class(TC) value [required]
--dot1p INTEGER RANGE Cos value
--remove Delete the mapping entry
-h, -?, --help Show this message and exit.
Step 3: Check the remark profile.
admin@sonic:~$ show qos remark dot1p
dot1p policy: remark_dot1p
TC Dot1p
---- -------
0 1
1 2
Step 4: Binding the remark table to the egress interface.
admin@sonic:~$ sudo config interface qos remark dot1p bind Ethernet8 remark_dot1p
Note:
admin@sonic:~$ sudo config interface qos remark dot1p --help
Usage: config interface qos remark dot1p [OPTIONS] <op> <interface_name>
<profile>
tc-dot1p policy configuration
Options:
-?, -h, --help Show this message and exit.
- "op": "bind/unbind"
Step 5: Check the binding table.
admin@sonic:~$ show interfaces qos
Ethernet8:
Dot1p remark: remark_dot1p
Result:
Since this example didn't set the QoS Classification, when the VLAN 10 packet with priority 0 into switch, it will base on the default QoS Classification.
The priority 0 will map to TC 0.
Base on the remark profile, the TC 0 will map to the Dot1p 1 which means the traffic through the TC 0 will be remarked to priority 1.
Here's the packet.
QoS Scheduler
Queue scheduling provides preferential treatment of traffic classes mapped to specific egress queues. SONiC supports SP, WRR, and DWRR scheduling disciplines.
- SP (Strict Priority)- Higher priority egress queues get scheduled for transmission over lower priority queues.
- WRR (Weighted Round Robin) - Egress queues receive bandwidth proportional to the configured weight. The scheduling granularity is per packet which causes large and small packets to be treated the same. Flows with large packets have an advantage over flows with smaller packets.
- DWRR (Deficit Weighted Round Robin) - Similar to WRR but uses deficit counter scheduling granularity to account for packet size variations and provide a more accurate proportion of bandwidth.
Tested model & firmware version:
- Switch model name:
DCS203 (AS7326-56X)
- Edgecore SONiC version:
202111.1
202111.3
Topology:
Pre-configuration:
Add VLAN 10 to Ethernet0, Ethernet1, and Ethernet5.
admin@sonic:~$ show vlan brief
+-----------+--------------+-----------+----------------+-------------+---------------+---------------+-------------+
| VLAN ID | IP Address | Ports | Port Tagging | Proxy ARP | DHCP Helper | DHCP Source | DHCP Link |
| | | | | | Address | Interface | Selection |
+===========+==============+===========+================+=============+===============+===============+=============+
| 10 | | Ethernet0 | tagged | disabled | | | |
| | | Ethernet1 | tagged | | | | |
| | | Ethernet5 | tagged | | | | |
+-----------+--------------+-----------+----------------+-------------+---------------+---------------+-------------+
The QoS Classification uses the default setting.
SP (Strict Priority)
Procedure:
Step 1: Set the scheduler mode.
admin@sonic:~$ sudo config scheduler add strict_mode --sched_type STRICT
Note:
admin@sonic:~$ sudo config scheduler add --help
Usage: config scheduler add [OPTIONS] <profile_name>
Add QoS-Scheduler profile.
Options:
--sched_type [WRR|DWRR|STRICT] Scheduler type. WRR is not supported on
Intel platform.
--weight INTEGER RANGE weight
--shaper_type [bytes|packets] Shaper type
--bandwidth TEXT Maximum bandwidth rate in kbps or pps. If
shaper-type is bytes, user can specify a
decimal number followed by the abbreviation
k (1000), m (1,000,000), or g
(1,000,000,000)
-h, -?, --help Show this message and exit.
Step 2: Check status.
admin@sonic:~$ show scheduler
Name Scheduling Type Weight Shaper Type Bandwidth
----------- ----------------- -------- ------------- -----------
strict_mode STRICT N/A N/A N/A
Step 3: Binding the scheduler to the interface.
admin@sonic:~$ sudo config interface scheduler bind queue Ethernet5 3 strict_mode
Note:
admin@sonic:~$ sudo config interface scheduler bind queue --help
Usage: config interface scheduler bind queue [OPTIONS] <interface_name>
<queue> <sched_policy_name>
Options:
-h, -?, --help Show this message and exit.
Here's the command to unbind the scheduler.
admin@sonic:~$ sudo config interface scheduler unbind queue Ethernet5 3
Step 4: Check status.
admin@sonic:~$ show interfaces scheduler
Ethernet5
Type Queue Profile Scheduling Mode Weight Shaper Type Bandwidth
------ ------- ----------- ----------------- -------- ------------- -----------
Queue 3 strict_mode STRICT N/A N/A 0
Result:
Injecting the traffic with priority 0~7 from Ethernet0 and Ethernet1 to Ethernet5 to trigger the congestion.
Queue 3 in Ethernet5 will be the highest priority.
admin@sonic:~$ show queue counters Ethernet5
Last cached time was 2022-09-14 06:40:04.548068
Port TxQ Counter/pkts Counter/bytes Drop/pkts Drop/bytes
--------- ----- -------------- --------------- ----------- ------------
Ethernet5 UC0 1,824,146 233,490,688 2,407,299 308,134,272
Ethernet5 UC1 1,824,176 233,494,528 2,407,270 308,130,560
Ethernet5 UC2 1,824,179 233,494,912 2,407,292 308,133,376
Ethernet5 UC3 4,255,417 544,693,376 0 0
Ethernet5 UC4 1,824,185 233,495,680 2,407,305 308,135,040
Ethernet5 UC5 1,824,187 233,495,936 2,407,276 308,131,328
Ethernet5 UC6 1,824,189 233,496,192 2,407,272 308,130,816
Ethernet5 UC7 1,824,194 233,496,832 2,407,284 308,132,352
Ethernet5 UC8 0 0 0 0
Ethernet5 UC9 0 0 0 0
Ethernet5 MC10 0 0 0 0
Ethernet5 MC11 0 0 0 0
Ethernet5 MC12 0 0 0 0
Ethernet5 MC13 0 0 0 0
Ethernet5 MC14 0 0 0 0
Ethernet5 MC15 0 0 0 0
Ethernet5 MC16 0 0 0 0
Ethernet5 MC17 0 0 0 0
Ethernet5 MC18 0 0 0 0
Ethernet5 MC19 0 0 0 0
Notes:
- 8 dedicated queues (i.e UC0, UC1, ..., UC7) for unicast.
- Other 8 dedicated queues (i.e MC10, MC11, ..., UC17) for multicast.
- UC8, UC9, UC18, UC19 are not available.
WRR (Weighted Round Robin)
Procedure:
Step 1: Set the scheduler mode.
admin@sonic:~$ sudo config scheduler add wrr_7 --sched_type WRR --weight 7
admin@sonic:~$ sudo config scheduler add wrr_3 --sched_type WRR --weight 3
Step 2: Check status.
admin@sonic:~$ show scheduler
Name Scheduling Type Weight Shaper Type Bandwidth
------ ----------------- -------- ------------- -----------
wrr_3 WRR 3 N/A N/A
wrr_7 WRR 7 N/A N/A
Step 3: Binding the scheduler to the interface.
admin@sonic:~$ sudo config interface scheduler bind queue Ethernet5 3 wrr_7
admin@sonic:~$ sudo config interface scheduler bind queue Ethernet5 4 wrr_3
Step 4: Check status.
admin@sonic:~$ show interfaces scheduler
Ethernet5
Type Queue Profile Scheduling Mode Weight Shaper Type Bandwidth
------ ------- --------- ----------------- -------- ------------- -----------
Queue 3 wrr_7 WRR 7 N/A 0
Queue 4 wrr_3 WRR 3 N/A 0
Result:
Injecting the traffic with priority 3,4 from Ethernet0 and Ethernet1 to Ethernet5 to trigger the congestion.
The egress traffic will be distributed based on weights.
admin@sonic:~$ show queue counters Ethernet5
Last cached time was 2022-09-23 05:53:03.667274
Port TxQ Counter/pkts Counter/bytes Drop/pkts Drop/bytes
--------- ----- -------------- --------------- ----------- ------------
Ethernet5 UC0 1 409 0 0
Ethernet5 UC1 0 0 0 0
Ethernet5 UC2 0 0 0 0
Ethernet5 UC3 1,518,614 858,150,900 620,044 348,958,020
Ethernet5 UC4 671,393 379,639,721 1,467,263 827,705,931
Ethernet5 UC5 0 0 0 0
Ethernet5 UC6 0 0 0 0
Ethernet5 UC7 0 0 0 0
Ethernet5 UC8 0 0 0 0
Ethernet5 UC9 0 0 0 0
Ethernet5 MC10 0 0 0 0
Ethernet5 MC11 0 0 0 0
Ethernet5 MC12 0 0 0 0
Ethernet5 MC13 0 0 0 0
Ethernet5 MC14 0 0 0 0
Ethernet5 MC15 0 0 0 0
Ethernet5 MC16 0 0 0 0
Ethernet5 MC17 0 0 0 0
Ethernet5 MC18 0 0 0 0
Ethernet5 MC19 0 0 0 0
Note:
Total: 1,518,614 + 671,393 = 2,190,007
Queue 3: 1,518,614 / 2,190,007 = 0.6934
Queue 4: 671,393 / 2,190,007 = 0.3065
DWRR (Deficit Weighted Round Robin)
Procedure:
Step 1: Set the scheduler mode.
admin@sonic:~$ sudo config scheduler add dwrr_7 --sched_type DWRR --weight 7
admin@sonic:~$ sudo config scheduler add dwrr_3 --sched_type DWRR --weight 3
Step 2: Check status.
admin@sonic:~$ show scheduler
Name Scheduling Type Weight Shaper Type Bandwidth
------ ----------------- -------- ------------- -----------
dwrr_3 DWRR 3 N/A N/A
dwrr_7 DWRR 7 N/A N/A
Step 3: Binding the scheduler to the interface.
admin@sonic:~$ sudo config interface scheduler bind queue Ethernet5 3 dwrr_7
admin@sonic:~$ sudo config interface scheduler bind queue Ethernet5 4 dwrr_3
Step 4: Check status.
admin@sonic:~$ show interfaces scheduler
Ethernet5
Type Queue Profile Scheduling Mode Weight Shaper Type Bandwidth
------ ------- --------- ----------------- -------- ------------- -----------
Queue 3 dwrr_7 DWRR 7 N/A 0
Queue 4 dwrr_3 DWRR 3 N/A 0
Result:
Injecting the traffic with priority 3,4 from Ethernet0 and Ethernet1 to Ethernet5 to trigger the congestion.
The egress traffic will be distributed based on weights.
admin@sonic:~$ show queue counters Ethernet5
Last cached time was 2022-09-28 07:07:32.975167
Port TxQ Counter/pkts Counter/bytes Drop/pkts Drop/bytes
--------- ----- -------------- --------------- ----------- ------------
Ethernet5 UC0 3 1,227 0 0
Ethernet5 UC1 0 0 0 0
Ethernet5 UC2 0 0 0 0
Ethernet5 UC3 5,972,663 764,500,864 2,473,283 316,580,224
Ethernet5 UC4 2,605,564 333,512,192 5,840,382 747,568,896
Ethernet5 UC5 0 0 0 0
Ethernet5 UC6 0 0 0 0
Ethernet5 UC7 0 0 0 0
Ethernet5 UC8 0 0 0 0
Ethernet5 UC9 0 0 0 0
Ethernet5 MC10 0 0 0 0
Ethernet5 MC11 0 0 0 0
Ethernet5 MC12 0 0 0 0
Ethernet5 MC13 0 0 0 0
Ethernet5 MC14 0 0 0 0
Ethernet5 MC15 0 0 0 0
Ethernet5 MC16 0 0 0 0
Ethernet5 MC17 0 0 0 0
Ethernet5 MC18 0 0 0 0
Ethernet5 MC19 0 0 0 0
Note:
Total:5,972,663 + 2,605,564 = 8,578,227
Queue 3: 5,972,412 / 8,577,474 = 0.6962
Queue 4: 2,605,062 / 8,577,474 = 0.3037
Hybrid mode (SP + WRR)
Procedure:
Step 1: Set the scheduler mode.
admin@sonic:~$ sudo config scheduler add strict_mode --sched_type STRICT
admin@sonic:~$ sudo config scheduler add wrr_7 --sched_type WRR --weight 7
admin@sonic:~$ sudo config scheduler add wrr_3 --sched_type WRR --weight 3
Step 2: Check status.
admin@sonic:~$ show scheduler
Name Scheduling Type Weight Shaper Type Bandwidth
----------- ----------------- -------- ------------- -----------
strict_mode STRICT N/A N/A N/A
wrr_3 WRR 3 N/A N/A
wrr_7 WRR 7 N/A N/A
Step 3: Binding the scheduler to the interface.
admin@sonic:~$ sudo config interface scheduler bind queue Ethernet5 2 strict_mode
admin@sonic:~$ sudo config interface scheduler bind queue Ethernet5 3 wrr_7
admin@sonic:~$ sudo config interface scheduler bind queue Ethernet5 4 wrr_3
Step 4: Check status.
admin@sonic:~$ show interfaces scheduler
Ethernet5
Type Queue Profile Scheduling Mode Weight Shaper Type Bandwidth
------ ------- ----------- ----------------- -------- ------------- -----------
Queue 2 strict_mode STRICT N/A N/A 0
Queue 3 wrr_7 WRR 7 N/A 0
Queue 4 wrr_3 WRR 3 N/A 0
Result:
Injecting the traffic with priority 2,3,4 from Ethernet0 and Ethernet1 to Ethernet5 to trigger the congestion.
Queue 2 in Ethernet5 will be the highest priority.
Queue 3, and 4 in Ethernet 5 will distribute the remaining traffic based on weights.
admin@sonic:~$ show queue counters Ethernet5
Last cached time was 2022-09-23 05:39:48.815061
Port TxQ Counter/pkts Counter/bytes Drop/pkts Drop/bytes
--------- ----- -------------- --------------- ----------- ------------
Ethernet5 UC0 1 409 0 0
Ethernet5 UC1 0 0 0 0
Ethernet5 UC2 5,630,631 720,720,768 0 0
Ethernet5 UC3 2,035,749 260,575,872 3,594,882 460,144,896
Ethernet5 UC4 912,664 116,820,992 4,717,966 603,899,648
Ethernet5 UC5 0 0 0 0
Ethernet5 UC6 0 0 0 0
Ethernet5 UC7 0 0 0 0
Ethernet5 UC8 0 0 0 0
Ethernet5 UC9 0 0 0 0
Ethernet5 MC10 0 0 0 0
Ethernet5 MC11 0 0 0 0
Ethernet5 MC12 0 0 0 0
Ethernet5 MC13 0 0 0 0
Ethernet5 MC14 0 0 0 0
Ethernet5 MC15 0 0 0 0
Ethernet5 MC16 0 0 0 0
Ethernet5 MC17 0 0 0 0
Ethernet5 MC18 0 0 0 0
Ethernet5 MC19 0 0 0 0
Note:
Total: 2,035,749 + 912,664 = 2,948,413
Queue 3: 2,035,749 / 2,948,413 = 0.6904
Queue 4: 912,664 / 2,948,413 = 0.3095
Hybrid mode (SP + DWRR)
Procedure:
Step 1: Set the scheduler mode.
admin@sonic:~$ sudo config scheduler add strict_mode --sched_type STRICT
admin@sonic:~$ sudo config scheduler add dwrr_7 --sched_type DWRR --weight 7
admin@sonic:~$ sudo config scheduler add dwrr_3 --sched_type DWRR --weight 3
Step 2: Check status.
admin@sonic:~$ show scheduler
Name Scheduling Type Weight Shaper Type Bandwidth
----------- ----------------- -------- ------------- -----------
dwrr_3 DWRR 3 N/A N/A
dwrr_7 DWRR 7 N/A N/A
strict_mode STRICT N/A N/A N/A
Step 3: Binding the scheduler to the interface.
admin@sonic:~$ sudo config interface scheduler bind queue Ethernet5 2 strict_mode
admin@sonic:~$ sudo config interface scheduler bind queue Ethernet5 3 dwrr_7
admin@sonic:~$ sudo config interface scheduler bind queue Ethernet5 4 dwrr_3
Step 4: Check status.
admin@sonic:~$ show interfaces scheduler
Ethernet5
Type Queue Profile Scheduling Mode Weight Shaper Type Bandwidth
------ ------- ----------- ----------------- -------- ------------- -----------
Queue 2 strict_mode STRICT N/A N/A 0
Queue 3 dwrr_7 DWRR 7 N/A 0
Queue 4 dwrr_3 DWRR 3 N/A 0
Result:
Injecting the traffic with priority 2,3,4 from Ethernet0 and Ethernet1 to Ethernet5 to trigger the congestion.
Queue 2 in Ethernet5 will be the highest priority.
Queue 3, and 4 in Ethernet 5 will distribute the remaining traffic based on weights.
admin@sonic:~$ show queue counters Ethernet5
Last cached time was 2022-09-29 02:36:49.112556
Port TxQ Counter/pkts Counter/bytes Drop/pkts Drop/bytes
--------- ----- -------------- --------------- ----------- ------------
Ethernet5 UC0 8 2,989 0 0
Ethernet5 UC1 0 0 0 0
Ethernet5 UC2 1,898,110 242,958,080 0 0
Ethernet5 UC3 1,180,033 667,767,452 718,077 403,717,686
Ethernet5 UC4 521,926 295,427,164 1,376,184 776,243,178
Ethernet5 UC5 0 0 0 0
Ethernet5 UC6 0 0 0 0
Ethernet5 UC7 0 0 0 0
Ethernet5 UC8 0 0 0 0
Ethernet5 UC9 0 0 0 0
Ethernet5 MC10 0 0 0 0
Ethernet5 MC11 0 0 0 0
Ethernet5 MC12 0 0 0 0
Ethernet5 MC13 0 0 0 0
Ethernet5 MC14 0 0 0 0
Ethernet5 MC15 0 0 0 0
Ethernet5 MC16 0 0 0 0
Ethernet5 MC17 0 0 0 0
Ethernet5 MC18 0 0 0 0
Ethernet5 MC19 0 0 0 0
Note:
Total: 1,180,033 + 521,926 = 1,701,959
Queue 3: 1,180,033 / 1,701,959 = 0.6933
Queue 4: 521,926 / 1,701,959 = 0.3066
QoS Shaping
Queue shaping provides control of minimum and maximum bandwidth requirements per egress queue for more effective bandwidth utilization. (Note: There is no minimum bandwidth guarantee support for queue shaping on Broadcom and Intel switches.) Egress queues that exceed an average transmission rate beyond the shaper max bandwidth will stop being serviced. Additional ingress traffic will continue to be stored on the egress queue until the queue size is exceeded which results in tail drop.
Tested model & firmware version:
- Switch model name:
DCS203 (AS7326-56X)
- Edgecore SONiC version:
202111.1
202111.3
Topology:
Pre-configuration:
Add VLAN 10 to Ethernet0, Ethernet1, Ethernet5.
admin@sonic:~$ show vlan brief
+-----------+--------------+-----------+----------------+-------------+---------------+---------------+-------------+
| VLAN ID | IP Address | Ports | Port Tagging | Proxy ARP | DHCP Helper | DHCP Source | DHCP Link |
| | | | | | Address | Interface | Selection |
+===========+==============+===========+================+=============+===============+===============+=============+
| 10 | | Ethernet0 | tagged | disabled | | | |
| | | Ethernet1 | tagged | | | | |
| | | Ethernet5 | tagged | | | | |
+-----------+--------------+-----------+----------------+-------------+---------------+---------------+-------------+
Procedure:
Step 1: Create the profile for QoS sharping.
admin@sonic:~$ sudo config scheduler add 400m_shaping --shaper_type bytes --bandwidth 400m
Note:
admin@sonic:~$ sudo config scheduler add --help
Usage: config scheduler add [OPTIONS] <profile_name>
Add QoS-Scheduler profile.
Options:
--sched_type [WRR|DWRR|STRICT] Scheduler type. WRR is not supported on
Intel platform.
--weight INTEGER RANGE weight
--shaper_type [bytes|packets] Shaper type
--bandwidth TEXT Maximum bandwidth rate in kbps or pps. If
shaper-type is bytes, user can specify a
decimal number followed by the abbreviation
k (1000), m (1,000,000), or g
(1,000,000,000)
-h, -?, --help Show this message and exit.
Step 2: Check the profile:
admin@sonic:~$ show scheduler
Name Scheduling Type Weight Shaper Type Bandwidth
------------ ----------------- -------- ------------- -----------
400m_shaping N/A N/A bytes 400 Mbps
Step 3: Binding the QoS sharping to the egress interface.
admin@sonic:~$ sudo config interface scheduler bind queue Ethernet5 3 400m_shaping
Note:
Usage: config interface scheduler <bind/unbind> queue [OPTIONS] <interface_name> <queue> <sched_policy_name>
Options:
-?, -h, --help Show this message and exit.
Or
Usage: config interface scheduler <bind/unbind> port [OPTIONS] <interface_name> <sched_policy_name>
Options:
-h, -?, --help Show this message and exit.
Step 4: Check the binding table.
admin@sonic:~$ show interfaces scheduler
Ethernet5
Type Queue Profile Scheduling Mode Weight Shaper Type Bandwidth
------ ------- ------------ ----------------- -------- ------------- -----------
Queue 3 400m_shaping N/A N/A bytes 400 Mbps
Result:
Since the QoS shaping, the traffic through to the Queue 3 will be down to about 400 Mbps.
Comments
1 comment
very nicely written article..simple and clean. Thanks
Please sign in to leave a comment.