Load balancing and high availability subsystem and method for distributed systems

Abstract

The present invention provides a load balancing and high availability subsystem and method for distributed systems, including at least two service provider modules, a registration center module and at least one client module; the service provider module calculates its own service weight , and register the services and service weights that it can provide to the registration center module, and calculate the service provider probability distribution of such services; the registration center module uniformly names the services registered by the service provider module in the structure of a naming tree service, and store the service provider list and service provider probability distribution corresponding to each service; the client module obtains the corresponding service provider list and service provider probability distribution according to the service name it wants to obtain, and uses the load balancing algorithm Randomly select a service provider; if the service provider is available, call the service module directly; if the service provider is not available, select an available service provider through a high availability algorithm.

Description

technical field [0001] The invention relates to the field of distributed systems, in particular to a load balancing and high availability subsystem and method for distributed systems. Background technique [0002] In most distributed systems, a large number of users access a certain resource concurrently. At this time, if the resource is concentrated on one server, it is obvious that this server will be under tremendous pressure. In this case, the CPU, memory, and I / O system of this server will soon reach the upper limit of performance, and eventually become the performance bottleneck of the system. Since a single server always has a performance upper limit no matter how its hardware configuration is improved, if the concurrent user scale continues to expand, a single server obviously cannot meet the ever-expanding concurrent scale. Only by adopting the server cluster method can the cluster scale be dynamically adjusted according to the concurrent scale effectively, and fin...

AI Technical Summary

Problems solved by technology

First of all, the schedule table is in the central control system, which means that the central control system needs to coordinate the execution of each client request; secondly, coordinating each client request consumes the resources of the central control system, which may lead to the emergence of bottlenecks; finally, When a server goes down or a new server is added, the schedule needs to consider the previous schedule to regenerate the schedule

Images