I'm just looking to get clarity on how exactly scheduling works on Cisco gear. According to a Cisco doc on scheduler operation:
HOW SCHEDULE ENTRIES ARE PROGRAMMED
The Min (minimum rate) entry allocates a minimum bvandwidth guaranteed amount of throughput to a queue. The Min entry is configured with the bandwidth command and is not set unless explicitly configured. IOS configuration checking attempts to ensure that a schedule will always have sufficient bandwidth to honor any configured Min rates. Servicing queues based on monitoring throughput vs. a preconfigured target rate is sometimes referred to as real time scheduling (refer to Scheduler's Representation of Time).
The Max (maximum rate) entry establishes a ceiling on the amount of throughput a queue can receive. The Max entry is configured using the shape command and is not set unless explicitly configured. Understand that Max sets a ceiling on the throughput of a queue but does not in itself guarantee any throughput to that queue.
The Ex (excess weight) entry mandates how queues will compete for any bandwidth available after Priority and Min guarantees have been met (excess bandwidth, or available bandwidth that is not guaranteed to, or not used by, priority and bandwidth guarantees). We configure Excess Weight with the bandwidth remaining command and unless explicitly configured, it defaults to 1. Excess bandwidth sharing is proportional to a queue's Excess Weight (sometimes referred to as virtual time scheduling, because no rates are configured and relative behavior alone is significant). For reflections on bandwidth sharing, see How Schedule Entries are Programmed.
SCHEDULE OPERATION
1 If the P1 queue is not empty, send the P1 packets.
2 If the PI queue is empty but the P2 queue is not, send the P2 packets.
3 Provided all priority queues are empty, the schedule services any queues with a minimum bandwidth
guarantee (Min) and continues to service such queues until the guarantees are met. To ensure fairness, the
scheduler will pick between queues with minimum guarantees by selecting the eligible queue, a queue
that has not exceeded the bandwidth guarantee and has been waiting longest.
4 What if priority queues are empty and all bandwidth guarantees have been satisfied? Any excess bandwidth
is distributed between queues that still require service until either all bandwidth is exhausted or a given
queue has reached a maximum configured bandwidth. The Ex configured in that queue's schedule entry,
dictates the share each queue will receive of this excess bandwidth.
So for all non-priority queues where a bandwidth statement is used to provide a minimum bandwidth guarantee:
- The scheduler chooses a queue that's still within the bw guarantee and has been waiting the longest.
- Transmit the packets from that queue until the guarantee is met or the queue is empty
- Moves on to the next queue that's still within the bw guarantee and has been waiting the longest
- After all the minimum bandwidth guarantees are met, the queues configured with bandwidth remaining are serviced. I assume that it also chooses the queue that's been waiting the longest.
After all the queues with allocated excess bandwidth are serviced, are excess bandwidth that's not allocated shared between all the non-priority queues? Is it done in a round-robin fashion? So if priority, bandwidth, and bandwidth remaining all add up to just 90% of the total bandwidth, would the remaining 10% be distributed between all non-priority queues? How is it distributed and how are the queues serviced?
Am I even on the right track or is my understanding way off?
