Fairness

Flow Rate Fairness : Dismantling a religion
  Internet has so many users that are trying to download and upload things at a time. Without a traffic control, the Internet would easily melt down. The traffic scheme we currently have focuses on the fairness of each flow. This paper criticizes the current scheme for its inefficiency and invalid notion of fairness. The paper argues that the current scheme is weak in attaining flow fairness moreover, it is not even achieving fairness in any sense. The author brings in an example that most applications use more than one flow at a time to speed up the connection and bring in the congestion. He mentions that it is so easy for people to violate the current congestion control avoidance and abuse the system with this kind of multiple flows cheat.
  Instead of trying to achieve fairness among each flow, the author argues that we should try to achieve fairness among each endpoint pretty much. In order to do so, we can leave everything else in current implementation as it is, and add an overarching connection supervisor to each machine. The goal is to have this supervisor regulate all the connections a machine is making and apply rate controlling according to weight parameters it would have for applications. He also wants to change the subscription plan of the ISPs to include costs for congestion and alleviation of congestion. Now then, users will be more mindful of the bits they are sending over the Internet because they would need to pay more to avoid congestion. The paper seems to have a pretty simple solution but I think that it is a waste of time to implement what this author is saying because this idea would bring more cost to use Internet and an unnecessary supervisor role in TCP/IP that can slow down the performance of the network.


Resource Pricing and the evolution of congestion control
  This paper goes along with the Flow Rate Fairness paper by Briscoe; in fact, this is the mother of the Flow Rate Fairness paper. It argues for TCP to use its congestion control mechanism to charge people for the congestion they are causing. For each marked packet, the user receives the ISP can charge some rate so that the users would control their traffic and thus balancing the workload of the Internet. It is interesting that the problem of the Internet is looked from almost the perspective of an economist. This paper's idea in a nugget is that an excess demand for bandwidth is going to be balanced by the increase in price.
  Through many scenarios and analysis, the paper shows that it is reasonable to have the users pay for what(congestion) they cause. People would have another incentive to abide by the TCP because they do not want to spend more money by overloading the network with many packets. It is a noble idea to worth consider especially now that even Internet browser starts to cheat on TCP by opening many simultaneous flows.
  However, the paper also concerns about the multicast issues that can charge users unknowingly. The nature of working with layers suggests that the user at TCP layer level won't know anything about the IP and link layer's work. The packets might get multicasted and be marked to bring up the cost the user has to pay. Certainly there is a danger in charging users for how many marked packets they get back. In ten years, it is likely that ISP companies will use this kind of economic solution to the congestion problem instead of fair sharing.