A method for implementing container load balancing

Abstract

The invention provides a high-performance and high-availability container load balance achieving method which adopts an application layer, a Tengine load layer, a LVS load layer and a registration center. The Tengine load layer and the LVS load layer form a double-layer distribution architecture, and rich load balance strategies are supported while the handling capacity of a whole cluster system is greatly improved. A unique and stable inlet is provided for a client-side through a virtual IP, and high availability of the load inlet is ensured. The application layer permits deployment of multiple applications, each application is configured with a matched application agent program. The Tengine load layer permits deployment of multiple Tengine, each Tengine is configured with one matched Tengine agent program. The LVS load layer permits deployment of multiple LVS, and each LVS is configured with one matched LVS agent program. Each layer of agent program is matched with the registration center to achieve serve registration capability, so that the whole architecture has transverse extending capability.

Description

technical field [0001] The invention relates to a container load balancing implementation method of a double-layer flow distribution architecture. Background technique [0002] With the popularity of Docker container technology, both the telecommunications industry and the Internet industry use Docker technology to run their applications in containers. For how to manage large-scale container online systems, there are three main solutions in the industry, namely Docker Swarm, Google Kubernetes and Apache Mesos. Among them, the most popular solution is Google's Kubernetes. The present invention will improve the load balancing of containers in Kubernetes, and propose a higher-performance and high-availability container load balancing implementation method. [0003] Kubernetes provides Service to provide reverse proxy and load balancing capabilities for multiple application containers in the backend, but this method provided by Kubernetes has some shortcomings: [0004] The ac...

AI Technical Summary

Problems solved by technology

[0005] Limited load balancing strategies lead to fewer application scenarios

The Service solution of Kubernetes only supports polling and IP_HASH (source address hash scheduling), and cannot be based on forwarding strategies such as URL / HEADER / COOKIE

[0006] Low load forwarding efficiency, not suitable for large-scale application clusters

The Service of Kubernetes implements request forwarding through NAT conversion of Iptables. For the request load of large-scale applications, this forwarding mode will become a performance bottleneck

Images