0

I'm building a network application and need to test its behavior under various bandwidths.

This question is an offshoot of this question. I was successfully able to test my application for different sets of bandwidth rates.

In addition I now want to test the application in which the bandwidth rate changes while the application is executing.

I figure that the same can be done manually/using a bash script (programmatically adding or deleting qdiscs), etc.

But it would be really useful if a tc command existed where I could simply provide a range of bandwidth values and the application switched to a random rate for the next packet transfer.

PS: I am very new to tc and therefore relatively verbose explanations / links to the same would be very helpful. Thanks!

Community
  • 1
spiritusozeans
  • 225
  • 2
  • 10

1 Answers1

0

The netem has most of the features required to simulate a physical network.

Specifically in this question, 

# tc qdisc add dev eth0 root netem delay 100ms 10ms 25%

causes the packets sent through interface eth0 to experience a delay of 100ms ± 10ms. Therefore the pings would produce a round trip time (rtt) ranging from 180ms to 220 ms. (2*delay)

The delay in different instances isn't completely random though. We may, if we so wish chose to have a correlation between consecutive values.

The value 25% denotes that the next delay value has a 25% dependency on the previous delay value.

This isn't true statistical correlation, but an approximation.

We may also chose to have our random delays to follow a particular distribution

 # tc qdisc change dev eth0 root netem delay 100ms 20ms distribution normal

Note: These examples has been borrowed from the netem tutorial which possesses many other useful examples as well.

spiritusozeans
  • 225
  • 2
  • 10