Hi guys, little question about Computer Networks regarding the Congestion Control in TCP Protocol. When does the congestion starts for the first time?
> Protip: It's being used the Reno Congestion Control algorithm.
>>59548853
Graph related
>>59548853
Do you take Network Engineering, anon? I don;t know the answer to your question but i'd like to hear your opinion on the course.
>>59548853
>When does the congestion starts for the first time?
It starts at 150 seconds
The Window size going from max to min indicates TCP slow start is happening. If it's happening on all of your flows you have a global sync problem and you should implement QOS or get better network devices
>>59548873
I studied a few years ago, good experience. We used Mininet, Python and Backtrack extensively.
>>59548933
Actually, I guess you could argue congestion starts at 120, because the two devices are negotiating for smaller window sizes.
Then at 150-155 someone failed to get an acknowledgement and the window size goes back to 1 and TCP slow start happens
>>59548964
I'm starting in September, looking forward to it and the prospect of a well paid job doing something interesting, should I learn anything before my course starts?
>>59548933
You're correct anon. What do you think it's hapening with the network at 120-140?
Do you think that the Reno algorithm is reacting well to the constant decrease of the lower bound threshold of the window?
> Protip: the decrease of the lower bound threshold at the interval might be related to the lack of ACK's of the segments sent at that particular interval (perhaps a problem in the link)
>>59548988
I really depends of the course you're taking, the metodology varies a lot. The common ground is Tanenbaum's book "Computer Networks". I suggest you play a little with some of the implemented algorithms. Mininet also offers you a enviroment to emulate a network. You can customize the topology using Python.
Good studies.
>>59548965
Yes, but at this moment the Reno algorithm is "exploring" how far it can explore the bandwidth until some segment timeouts. The problem starts here, but the algorithm probably hasn't detected the damn problem.
>>59549056
Come on guys, you're at one step in getting into the a wild trip to the rabbit hole.
:)
Time is not your allied
>>59548853
Looks more like CUBIC
>>59549723
The slow-start curve of the CUBIC algorithm is very discretized. Not like this graph.