公开(公告)号：JP2000353103A

公开(公告)日：2000-12-19

申请号：JP2000126482

申请日：2000-04-26

Applicant: IBM

Inventor： YOCUM PETER B ,  EILERT CATHERINE K ,  ARWE JOHN E

IPC: G06F9/46 ,  G06F9/50 ,  G06F15/177 ,  H04L12/28

Abstract: PROBLEM TO BE SOLVED: To enable the system management of the number of servers over clusters of a system by including a step for starting a server to a queue when it is decided that an operation request which has an affinity for only a subset of clusters having no server to a queue is present in the queue. SOLUTION: Clusters 90 of a computer system 100 which are connected to each other and cooperate are included. The environment in this case includes a queue 161 of operation requests 162 and a pool of servers 163 which are decentralized over the clusters 90 and services the operation requests. It is made possible to manage the number of the servers 163 according to the performance target classes of operations held in the queue and the performance target classes of operation conflicting in the computer system 100. The possession of one policy for the clusters 90 of the computer system 100 is useful to obtain a single image view of decentralized operation loads.

公开(公告)号：JPH11282695A

公开(公告)日：1999-10-15

申请号：JP3520599

申请日：1999-02-15

Applicant: Internatl Business Mach Corp ,  インターナショナル・ビジネス・マシーンズ・コーポレイション

Inventor： YOCUM PETER B ,  EILERT CATHERINE K ,  ARWE JOHN E

IPC: G06F9/46 ,  G06F9/50 ,  G06F15/177

CPC classification number: G06F9/5061 ,  G06F9/5083 ,  G06F2209/505

Abstract: PROBLEM TO BE SOLVED: To control the number of servers in a multisystem cluster.
SOLUTION: A computer system 100 has a queue 161 for organizing an inputted work request as a service class, transferring respective service classes to a cluster and serving them by servers 163. Prescribed performance indexes are previously allocated to respective service classes. Each system 100 selects a certain service class as a donor class for providing a system resource based on a degree of coincidence of each service class with its own target and selects another service class as a receiver class for receiving the system resource. When a resource bottleneck causing the receiver class to lose its target is the number of servers, each system 100 determines whether several servers are to be added to the receiver class or not based on judgement whether a plus effect of server addition to the performance index of the receiver class exceeds the minus effect applied to the performance index of a donor class or not.
COPYRIGHT: (C)1999,JPO

Abstract translation: 要解决的问题：控制多系统集群中的服务器数量。 解决方案：计算机系统100具有用于将输入的工作请求组织为服务类的队列161，将各个服务类别传送到群集并由服务器163提供服务​​。规定的性能指数预先分配给各个服务等级。 每个系统100基于每个服务类与其自己的目标的一致程度来提供系统资源的供体类别，选择某一服务类，并选择另一服务类作为用于接收系统资源的接收器类。 当导致接收者类别失去其目标的资源瓶颈是服务器数量时，每个系统100基于判断是否将服务器添加到性能指标的加号效应来确定是否将几个服务器添加到接收者类中 接收者类别超过了对捐赠者类别的绩效指标的负面效应。