0

I need some clarifying on that issue (little bit rusty on that)

Assuming I have some limitation on the BW of 1Mb. What will happen then there is incoming traffic of 1.5Mb for 10 sec - some of the data will start to arrive with delay? any other affect?

Thank you

  • 2
    Depends if the traffic shaping allows bursting of the traffic and if the total amount of bandwidth available can handle the requested 1.5Mb for 10 seconds. – user36472 Feb 10 at 8:34
1

This entirely depends on how the traffic limitation is implemented.

One way to achieve it is to have a link with 1Mb/s speed. Then, no data will arrive faster than 1Mb/s.

However, modern links are often faster, and the traffic shaping has been implemented in software. Then you should take a hard look at token bucket and leaky bucket. Chances are the rate limiting uses this kind of an algorithm.

Usually, there are buffers, no matter whether the rate limiting is link speed based or implemented in software. How large the buffers are is then another question. Many systems have excessive buffering (bufferbloat).

So, any of these can happen:

  1. The burst will be immediately allowed in as it's so small (allowing momentary bursts), but if you consider e.g. a 100-second burst, then the burst won't arrive immediately
  2. The burst will go through 1Mb/s link or software rate limitation, meaning the packets will arrive but later (delay)
  3. The burst is larger than buffer size, therefore some packets will be lost (loss)

You might think that (2) is better than (3), but the contrary is true. A 10 second buffer is way too large. Nobody likes a link with 10 second latency. It's usually better to have low latency than it's to have low loss.

Related:

1

Limitations can be physical (limit of a link) or logical (artificial limiting of a faster link by policy). Physical limits can't be exceeded.

With a logical limit, everything depends on how the policy is implemented in a device. It may allow the actual bandwidth to exceed the limit for a (very) short time (bursting).

Excess traffic is usually queued, but if the queue capacity gets exhausted, data overflows and is dropped.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.