主动自调度集群的自调度算法  

Automatic Self-Allocating Strategy of ASAS Cluster

作　　者：刘海[1] 王晓川[2] 金士尧[2] 

机构地区：[1]广东科学技术职业学院计算机工程技术学院,珠海519090 [2]国防科技大学计算机学院,长沙410073

出　　处：《现代计算机》2007年第5期7-10,共4页Modern Computer

基　　金：国家专利项(专利号:ZL02114011.1)

摘　　要：ASAS集群采用了基于后端服务器负载分析后的主动调度策略,在后端主动调度时计算后端服务器的负载,包括CPU利用率、内存利用率、I/O速度等,然后把上述几个方面划分为底(L)、中(M)、高(L)三个语言变量后,通过模糊数学理论,选取适当隶属函数,把实时状态隶属于以上三个语言变量的隶属函数值求出,并得出隶属矩阵。根据合适的算法得出后端执行服务器申请的连接数(容量Capacity),整个过程是基于负载动态反馈的。实验证明,该调度策略取得了比一般集群调度策略更优的效果。ASAS cluster depends on its back-scheduler-severs automatic self-allocating based on load status. Among the back-scheduler-server admeasure, firstly, calculates the load of the backscheduler-server which includes CPU usage ratio, memory usage ratio and I/0 speed etc. Secondly, with Fuzzy Mathematics, the top aspect （CPU usage rate, memory usage rate and I/O speed etc.）is divided into three language variable： Low, Middle, High. And then selects appropriate subject function and calculates the values of the real status subjected to language variable and achieves the subject matrix. Lastly, calculates the number of connection （Capacity） with appropriate method. The whole process is based on load active feedback. Conclusion can be drawn from. some experiments that allocation strategy of ASAS is better than others.

关 键 词：主动自调度 集群 隶属函数 隶属矩阵 负载 容量 

分 类 号：TP338[自动化与计算机技术—计算机系统结构]