Traffic Shaping Exceed Burst Size and Queue-limit - Confusion

[hsien-ping tang]

Hi Experts,

I am confused about the relationship between shape burst size and queue-limit.

I check the configuration guide, there are 2 locations can set the buffer size,
1. shpae with Excess burst size
2. queue-limit

If there is a burst traffic going to exit outbound interface, and the traffic is exceeded the shape rate but not over the PIR.
my question is where will those traffic be buffered?

What's the relationship between these two?

Thank you,

Tom

[smilstea · Accepted solution]

Think of your QoS policy like a pizza, your pizza is worth 1Gbps, queue-limit defines time and bandwidth for queueing packets and transmitting packets, IE you have 1 pizza, but you cut it into 10 equal slices, you (the interface) can only reasonably eat one slice at a time. Now lets say you are getting drops in QoS due to bursty traffic, you can increase your queue-limit or in this case cut the pizza into 6 slices, but it takes you longer to consume each slice but handles burstiness better (back to QoS). Burst size in shaping is how far above your committed rate you can go in case there is a burst, for instance on a 1Gbps policy you might be able to burst 100Mbps in excess.

So both concepts help with burstiness of traffic, just slightly differently, queue-limit expects your traffic over time, say 100ms increments to all be uniform, so if you burst 1Gbps in the first 100ms then have nothing in the next 900ms then you will drop 900Mbps, if your queue-limit is 200ms then you would drop 800Mbps but it takes longer to process and send the data.

Sam