Applications

When configuring LB for your services, you should select a proper traffic distribution algorithm to distribute traffic, which could avoid uneven server loads and protect service performance from being affected by overload on partial servers.

Algorithm Principle

Algorithm	Weighted round robin Weighted Round-Robin Scheduling	Least-connection scheduling	Source hashing scheduling (ip_hash)
Principles	The weighted round-robin scheduling algorithm is used to schedule requests to different servers based on polling. It can solve problems with imbalanced performance of different servers. It uses weight to represent the processing performance of a server and schedules requests to different servers by weight in a polling manner. Servers with a higher weight receive connections earlier and process more connections. Servers with the same weight will process the same number of connections.	In actual situations, the staying time of client requests on the server may vary greatly. As the working time gets longer, if a simple round-robin or random LB algorithm is used, the number of connection processes on each server may vary hugely, failing to achieve LB. Least-connection scheduling is a dynamic scheduling algorithm that estimates the load of a server by the number of active connections. Its working principle is contrary to round-robin scheduling. The scheduler must record the number of established connections on each server. If a request is scheduled to a server, its number of connections will be increased by 1. If a connection stops or times out, the number of connections will be decreased by 1.	It uses the source IP address of the request as the hash key and locates the corresponding server from the statically assigned hash table. The request will be delivered to this server if it is available and not overloaded; otherwise, null will be returned.
Advantages	The round robin algorithm is simple and highly practical. It is not required to record the status of all connections and is therefore a stateless scheduling algorithm.	It is applicable for requests requiring long-time processing, such as FTP.	ip_hash can realize partial session persistence, remember the source IP, and map requests from a client to the same backend host through the hash table.
Disadvantage	The weighted round-robin scheduling algorithm is relatively simple and is unsuitable for situations where the service time of a request changes significantly, or each request consumes different amounts of time. In these cases, it will cause imbalanced load distribution among servers.	For short connection services with frequently changing number of server connections. it does not have obvious advantage.	In scenarios with large difference in traffic volume, it will easily lead to uneven load.
Applicable Scenarios	Each request consumes basically the same amount of backend time with the best loading performance. It is usually used in short connection services such as HTTP service.	The time used by each request on the backend varies greatly. It is usually used in long connection services.	It is suitable for LB of TCP/UDP protocols. ip_hash scheduling can be used in scenarios where session persistence is not supported.

Precondition

A VPC has been created in the availability zone, and a business host has been created with their service configurations completed.
A load balance instance has been created.

Operation Steps

Click the ELB instance name to enter the listener configuration page.
On the listener page, click Add Listener to create a listener.
On the listener creation page, enter a listener name, select a protocol and enter a listener port and description. Click Next.
Create a backend host group, select a backend cloud host and add it to the backend host group, and then click Next;
Go to the LB and Health Checkconfiguration page, select a LB algorithm from the load mode field, and complete other configurations such as session persistence and health check.
Click Create Now. A listener is created and configured.

Elastic Load Balancing

Configuring the LB Algorithm