Technical Selection of Message Queuing

Message Queuing Enterprise Application Scenario Analysis

1. Message Queue: A container for storing messages during message transmission. Producers and consumers do not communicate directly, relying on queues to ensure message reliability and avoid mutual influence between systems.
2. Major roles of message queue: server, client
3. The main functions of message queues: business decoupling, asynchronous calls, traffic peak shaving
   • Advantages of decoupling: improve system stability, avoid multiple calls by broadcasting messages, and asynchronously process through message queues for scenarios where you don't need to pay attention to call results

Comparative analysis of message queue selection, why use RocketMQ

feature comparison	Maturity	Sequence Message	Delayed message	Transaction message	Message filtering	Message query	Consumption failure retry	Batch send
kafka	Log Field	Support	Not supported	Not supported	Support	Not supported	Not supported	Support
RabbitMQ	Mature	Support	Not supported	Not supported	Not supported	Not supported	Support	Not supported
RocketMQ	Mature	Support	Specific LV	Support	Support	Support	Support	Not supported

Utility	Deployment method	Cluster Management	Select the main method	Master-slave switch	Data reliability	Performance	Availability	Stackability
kafka	Clusters	Zookeeper	Automatic elections	Auto switch	High	Very high	Distributed, master-slave	Very good
RabbitMQ	Clusters	Erlang native support	The oldest joined node	Auto switch	High	in	Master-Slave	General
RocketMQ	Clusters	Name Server	Not supported	Auto switching is not supported	High	High	Distributed, master-slave	Very good

• Kafka: a data stream channel between systems
• RabbitMQ: Reliable Messaging
• RocketMQ: High-performance reliable messaging

Analysis of MQ High Availability and High Expansion Solutions

1. RocketMQ Architecture
   • Broker master-slave deployment, its own information is registered in NameServer
   • Client gets broker information from NameServer
   • NameServer nodes are independent of each other, no data interaction
2. Reliability Analysis
   • Sync Flash: low performance, high reliability
   • Asynchronous flushing: high performance, low reliability

• Asynchronous replication
• Sync Double Write

3. Usability Analysis
   • Master-slave mode Master is down: Broker is readable but not writable
   • How to build a cluster
     • Single Master Mode
     • Multi-Master Mode
     • Multi-Master Multi-Slave Mode-Asynchronous Replication
     • Multi-Master Multi-Slave Mode-Synchronous Double Write

Analysis of MQ service registration and discovery principles

1. Registry Center—NameServer Design
   • Service registration: Broker registers itself with information
   • Service Discovery: Procucer/Consumer Find Broker Information
2. Registry Center—NameServer Design Advantages
   • Design goals are simple and efficient
   • High availability solution: multiple nodes independent of each other
   • Data Storage Design
     • Cluster-
     • Broker1 -
     • Topic-
     • Broker-heartbeat