background

Large-scale e-commerce websites are based on SOA architecture. Such a large-scale service architecture requires an efficient, fast, and elegant service governance platform. This article introduces how to build an efficient service governance platform.

content

Service Governance Platform

The original intention of the service governance platform is to:

It can realize the management of service health status and track the whole life cycle of each service request

It can realize fault isolation, graceful degradation, quick response and localization of problems

Manage dependencies between services

When designing this service governance platform, several design principles were adopted:

distributed architecture
no center
no single point

Platform architecture design

Service Governance Platform Architecture

As shown in the figure, this is the architectural design of the service governance platform, which is built using open source technologies such as ZooKeeper, Detector, message middleware, MySQL, and MongoDB.

work process:

Step 1: The service provider (Service Provider) must first submit a registration application to the ZooKeeper Cluster, and only after the registration is successful can it provide external services.
Step 2: ZooKeeper Cluster pushes the list of available service providers to the service consumer, and the service consumer can only use the certified service provider in the list.
Step 3: The service user requests the service from the service provider.
Step 4: The service provider successfully responds to the service requester's request.

Other processes:

At the same time, the service provider and service user will push a call log to the Jumper Broker. The main content of the information is the call frequency, response time, etc. After the Jumper Broker analyzes and processes the information, it sends the result to the Detector.
Detector records the information and pushes the information to ZooKeeper Cluster.
If the response time of a service is getting slower and slower, ZooKeeper Cluster will find out and make adjustments in time, for example, no longer allocate so many calls to this service until its status returns to normal.

Other technical realization

From the figure above, it can be noticed that ZooKeeper Cluster, Jumper Broker, and Detector are all deployed in clusters, ensuring the high availability of the service governance platform itself.
In terms of technical implementation, the asynchronous message mechanism and RPC framework are also adopted, so that the architecture itself has no center and no single point, and can support tens of thousands of nodes.

The entire platform is also very simple to deploy, as long as the client of the service management platform is deployed together with the service, and some simple configurations are done.