This article summarizes the behavior of MX serial QOS, including Q card and MQ card.

How do you calculate the bandwidth for each queue, presuming the bandwidth is 10G?

1. Q1 20% low traffic 5G
   Q2 30% low traffic 5G
   Q3 straight-high traffic 5G


2. Q1 20% low traffic 5G
   Q2 30% low traffic 5G
   Q3 straight-high traffic 3G


3. Q1 10% low traffic 5G
   Q2 40% low traffic 5G
   Q3 straight-high traffic 6G


4. Q1 20% low traffic 5G
   Q2 30% medium-low traffic 5G
   Q3 straight-high traffic 6G

1. Q1 2G
   Q2 3G
   Q3 5G
   
   Summary: All traffic in straight-high is guarantee traffic, so 10-5 = 5G. Q1 and Q2 request 50% of 10G. Therefore, 5G will do WRR to allocate to each queue.


2. Q1 2.8G
   Q2 4.2G
   Q3 3G
   
   Summary: The queue 3 is straight high, so it will take the whole 3G. The leftover is 7G. It can satisfy all transmit-rate of Q1 and Q2. So Q1 will get 7*20/(20+30)=2.8. Q1 is 4.2G


3. Q1  1G
   Q2  3G
   Q3  6G
   
   Summary: Queue 3 takes 6G. The leftover is 4G. 4G cannot satisfy Q1 and Q2, so it will do RR fashion. Q1 gets 2G and Q2 gets 2G. But from the configuration, Q1 just need a 10% transmit rate. So Q1 just has 1G. The leftover goes into Q2.


4. Q1 1G
   Q2 3G
   Q3 6G
   
   Summary: Queue 3 takes 6G. The leftover is 4G. Because Q2 is medium queue, it will satisfy medium first. Q2 gets 3G. The leftover 1G allocates to Q1.

According to the results:

1. All straight traffic is guarantee traffic.
2. The bandwidth satisfies straight first, then the order is high, medium, low. 
3. If the leftover bandwidth cannot satisfy all transmit-rate of the same queue priority, it will do RR instead of WRR.