For each paper, by 12AM the evening before lecture:
1. What happens to the drop probability in PIE if the queueing delay is larger than (say, twice) the target delay (QDELAY_REF) for a very long time?
2. If you have a collection of TCP connections from different sources concurrently sharing a bottleneck link whose router runs PIE, would each connection get the same throughput?
2. Recall the TCP throughput formula. Assuming the same RTTs for the connections, the connections will have the same throughput (on average) because the probability of packet drop for each flow is the same.
How does ECMP work? What is it used for and what are its disadvantages?
Why does VL2 require TCP congestion control to achieve its goal of uniform high throughput and performance isolation? What can go wrong with non-TCP traffic?