RPC (Remote Procedure Call) protocol based cluster resource unified management system

Abstract

The invention relates to an RPC (Remote Procedure Call) protocol based cluster resource unified management system. According to the RPC protocol based cluster resource unified management system, a cluster system is divided into a host node and a plurality of slave nodes; a global resource management service is deployed to the host node; node management services and specific application program main control services aiming at different application programs are deployed to the slave nodes; resources on the slave nodes are packaged into resource containers; the resource management service is responsible for resource management and distribution of the integral system; the node services are responsible for management of resources and tasks on every node; the application program service main control services are responsible for management of the single application program. The RPC protocol based cluster resource unified management system which is a distribution system for main or auxiliary architectures integrally has the advantages of providing unified service ports for other distribution application programs, integrating a plurality of original distribution applications which operate independently and distribute cluster resources, enabling cluster software and hardware resources to be shared, improving the resource utilization rate, reducing operation and maintenance costs, achieving data sharing and enhancing expansibility.

Description

technical field [0001] The invention relates to a cluster resource unified management system based on the RPC protocol, and belongs to the technical field of cluster resource management. Background technique [0002] At present, there are many distributed application resource management systems, which can manage and schedule the resources required by jobs according to the characteristics of their respective systems. For example, Hadoop, an open source project of the Apache Open Source Software Foundation, only supports MapReduce jobs. The new scheduling strategy is difficult to embed in its central scheduler, and its scalability is poor. At the same time, the central scheduler may have a single point of failure and cause the failure of the entire system. As another example, Google optimizes the central scheduler and puts each scheduling strategy into a separate module. Different jobs are scheduled by different scheduling strategies. This solution can adapt to various types ...

AI Technical Summary

Problems solved by technology

For example, Hadoop, an open source project of the Apache Open Source Software Foundation, only supports MapReduce jobs, and the new scheduling strategy is difficult to embed in its central scheduler, with poor scalability

At the same time, the central scheduler may have a single point of failure and cause the entire system to fail

As another example, Google optimizes the central scheduler and puts each scheduling strategy into a separate module. Different jobs are scheduled by different scheduling strategies. This solution can adapt to various types of distributed applications. , but since all scheduling strategies are still in a centralized component, the scalability of the whole system has not become better

Images