Utility Based Congestion Control for Delay Constrained Communication

In network congestion control theory,contrary to Marxism,rules can be created and destroyed.

 this blog mostly is based on [1].
 As pointed by many network researchers,the delay based congestion control algorithms may get starvation when completing bandwidth resource with loss based congestion control algorithms.
 The utility function introduced first into network congestion control by Kelly,indeed is an efficient theory to create new congestion control algorithms. Because of the large scale of network,the centralized solution to the NUM problem is impractical. Fortunately,the problem can be solved in distributed way using a primal-dual algorithm.
$\dot x_r=\alpha_r\left( U'(x_r)-\sum\limits_{l\in L}R_{lr}\lambda_{l}\right)\tag{14a}$
$\dot\lambda_{l}=v_{l}(y_{l}-c_{l})_{\lambda_{l}}^+\tag{14b}$.
where$\ \lambda_{l}\ $can be treated as shadow price of link L such as queue delay and loss rate.In the case of queue delay as price,Extrapolate the precise of queue delay from total delay is challenging.And the delay based flow is not robust again loss based flow.
 So,given the challenges mentioned above,The author proposes the rate control law as follow:
$\dot x_t(t)=k_rx_r\left(U'(x_r)-v_r(e_r)-\dot e_r-\pi_{r}\right)_{x_r}^{+}\tag{15}$
 [2]提出基于用户的拥塞控制机制，一个用户可能会有多条连接，但是通过接入控制的方式，保证网络资源分配的公平性。
[1]Improved Utility-based Congestion Control for Delay-Constrained Communication.(2007-ton)
[2]Network Resource Allocation for Users With Multiple Connections: Fairness and Stability(2014-ton)