FP4 forwarding planes support the reallocation of queues between ingress and egress. Queues can be reallocated either from ingress to egress, or from egress to ingress. The reallocation is performed in sets of 8k (8192) queues, with the requirement that a minimum of 8k queues must be allocated at both ingress and egress. The ingress and egress buffer pool sizes are not affected by the queue reallocation.
The queue allocation is configured using a named FP resource policy that is created using the fp-resource-policy command within the QoS configuration branch and is applied under the card>fp context. The system supports up to 15 policies, including the system-created default policy that cannot be modified or deleted and is applied by default to all FP4 forwarding planes. The default policy allocates 128k queues to ingress and 128k queues to egress.
The queue allocation within the FP resource policy is configured as follows:
The ingress-percent-of-total command specifies the percentage of the total number of queues allocated to ingress queues, with the remainder being allocated to egress queues. The percentage is configured as an integer to two decimal places with a range from 4.00 to 97.00 (values outside of this range are not applicable due to the requirement of a minimum of 8k queues are ingress and egress).
Queues can be allocated in sets of 8k queues. If the percentage configured results in the queue allocation not being a multiple of 8k, the number of queues at ingress is rounded down to the next 8k boundary, and consequently the number of queues at egress is rounded up to the next 8192 boundary, both while respecting the minimum at ingress and egress. For example, if the ingress-percent-of-total is configured to 27%, that calculates to 70778.88 queues, the number of ingress queues would be rounded down to 65536 queues and the number of egress queues rounded up to 196608.
FP resource policies are applied to a forwarding plane as follows:
An FP resource policy that is applied to an FP cannot be deleted.
The queue allocation can be modified by either applying a differently named FP resource policy to an FP, and this includes removing an applied user-created FP resource policy to return to the default policy, or modifying the ingress-percent-of-total command that also includes removing this parameter to return to its default configuration in the policy applied to an FP. The system then checks whether requested queue allocation is achievable based on the current ingress and egress queue consumption on the related FP. If the requested queue allocation is not achievable, the command fails.
When the queue allocation is modified, the system immediately resets the associated cards, XIOMs, and MDAs in order to reallocate the queues, with the following exceptions:
When redundancy mechanisms are used that spans FPs, for example LAGs, it is recommended that the same queue allocation is applied to all related FPs.
FP resource policies can be displayed using a show command, with the details of the configured ingress-percent-of-total command and the associations on which FPs it is applied. The name of the FP resource policy applied to an FP can be displayed using the show card slot-number detail command. For example, the FP resource policy below allocates the minimum number of queues to ingress and the maximum number of queues to egress, and is applied to card 1 FP 1:
When an FP resources policy is applied to change the queue allocation, the resulting queue allocation can be displayed using the tools dump resource-usage card slot-num fp fp-number command: