具有差异化服务的混合对等网络性能分析  

Performance analysis of hybrid peer-to-peer networks with differentiated services

作　　者：张长振 马占友[1] 刘琳 陈利[2] ZHANG Changzhen;MA Zhanyou;LIU Lin;CHEN Li(School of Science,Yanshan University,Qinhuangdao 066004,China;LiRen College,Yanshan University,Qinhuangdao 066004,China)

机构地区：[1]燕山大学理学院,河北秦皇岛066004 [2]燕山大学里仁学院,河北秦皇岛066004

出　　处：《系统工程与电子技术》2023年第1期271-278,共8页Systems Engineering and Electronics

基　　金：国家自然科学基金(61973261,61872311);河北省自然科学基金(A2020203010,A2018203088)资助课题。

摘　　要：为了应对对等网络中的“搭便车”行为,提出一个惩罚策略,即对“搭便车”行为的请求节点和资源共享的请求节点提供差异化服务,减少“搭便车”行为的请求节点的个人收益。使用带有两类服务台、负顾客的多重工作休假M/M/c+d排队模型对混合对等网络进行建模,利用矩阵几何解的方法,得到了系统的稳态分布。通过数值实验定量刻画了参数变化对系统性能的影响,对比了请求节点的个人平均收益。实验结果显示,无论在哪一个休假参数水平下,Ⅰ类请求节点的个人平均收益相比于Ⅱ类请求节点的个人平均收益至少高出20%。A penalty strategy is proposed to deal with“free riding”behavior in peer-to-peer networks by providing differentiated services to requesting nodes for“free riding”behavior and requesting nodes for resource sharing to reduce the personal gain of requesting nodes for“free riding”behavior.We model the hybrid peer-to-peer networks using a multiple working vacation M/M/c+d queue model with two types of servers and negative customers.The stationary distribution of the system is obtained by using the method of matrix-geometric solution.Numerical experiments are conducted to quantitatively portray the impact of parameter variations on system performance and to compare the average individual revenue of requesting nodes.The experimental results show that the average individual revenue of type I requesting nodes is at least 20%higher compared to the average individual revenue of type II requesting nodes,regardless of the vacation parameter level.

关 键 词：混合对等网络 搭便车 排队模型 矩阵几何解 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]