简要:面向content providers(CPs);minmize the expected missed cache rate
1、属性
Content:
(1)、按种类划分等级;
(2)、流行度(被请求次数);
(3)、reserve slot expire at rate
cell station:
(1)、存储方面:
cp1及其他cp,自身需求
(2)、分布密度
2、Algorithm
(1)、Karush Kuhn Tucker(KKT) http://blog.csdn.net/oscarriddle/article/details/78790135
(2)Waterfilling solution
补:Finally, competition for the shared caching memory is formulated as a convex n–persons game: CPs trade off the expected missed cache rate for the memory price. It is a non-smooth Kelly mechanism with reservation and bounded strategy set – for which new existence and uniqueness of Nash equilibrium are provided. When CPs are myopic optimizers, convergence to the Nash equilibrium is showed numerically. Furthermore, the game appears to have a unique Stackelberg equilibrium, a relevant feature for the MNO in order to maximize her revenue. Online learning of the optimal price over time will be part of future works.
(3)、
Fir. 凸集:对于一个集合S中任意两个点A和B,这另个点的连线AB也在S内。
仿射函数:
Sec. 目标函数和不等式约束函数均为凸函数,等式约束函数为仿射和拿书,并且定义域为凸集的优化问题为带约束凸优化问题。
Thir. 解决带约束凸优化问题,牛顿迭代法
https://zhuanlan.zhihu.com/p/26514613
3、算法
gc:内容请求次数
在面积为pai r 的平方内未发现content的概率为
A为基站的分布密度,
剩下部分第i等级的内容。。。。
问题:指数一堆放一起?
未考虑带宽及user的移动速度(车联网是要考虑的);
考虑了不同内容不同等级以及不同等级在总数据数量的占有比例 && 不同CPs在基站的内存占有率 && CP i对不同等级的内容在基站缓存内的分配按权值分配
论文将问题转化为带约束凸优化问题,病用相关方法解决。
4、其他
网友评论