review: Charge-Sensitive TCP and Rate Control in the Internet
- What is the best way of
allocating the limited resources of the internet in a non-cooperative
environment? Is there a way of
creating a pricing scheme that will not create too much additional
overhead? What is necessary in
terms of fairness, especially when different streams are using different
- A proposal to solve the
distribution of resources while creating a pricing scheme and continuing
to ensure fairness.
- A. The problem of achieving the system
optimum is composed into two subproblems:
The optimum is achieved when users’ choices of charges and
the network’s choice of allocated rates are in equilibrium.
- To maximize the objective
function given the current prices per unit flow, a user only needs to know
the price per unit flow. Users with
TCP can computer their prices per unit flow without any help from the
network. They do not need to know
the price at each resource.
- Fairness is guaranteed
because the price that a user pays for using resources is proportional to
- As the number of users
increases, the rates at Nash equilibrium points converge to the system
- Critique the main
System researchers and
builders should recognize that the term fairness, even though not defined
in the non-cost TCP world, is even more ambiguous in the pricing rate
world. How we should charge people
and what should be considered fair is an issue that may not be
solvable. This seems to reduce the
issue of allowing each user to maximize their flows according to the
- Significance- 3 Although the
integration of pricing and fairness is seen as a step forward many of the
details are still in a theoretic stage.
I don’t see this being implemented without further studies.
- Convincing- 2 .The
evidence is largely theory based. The
equations seem to work correctly but have not been tested in a real
environment. They limited their
analysis to a simple single bottleneck.
It is extremely questionable if a user could predict the effect of
their parameters in a larger and more complicated network.