How ELB Works

Overview

Figure 1 How ELB works

The following describes how ELB works:

A client send requests to your application.
Each listener added to your load balancer uses the protocol and port you have configured to receive the requests.
The load balancer forwards requests.
1. Each listener forwards the requests to the associated backend server group based on the routing rules you configure.
2. If there are forwarding policies, each listener evaluates if incoming requests match any forwarding policy. If a match is found, the request is forwarded according to the forwarding action.
Healthy backend servers in the backend server group receive the requests based on the load balancing algorithm and the routing rules you specify in the forwarding policy, handle the requests, and return results to the client.

How requests are routed depends on the load balancing algorithms configured for each backend server group. If the listener uses HTTP or HTTPS, how requests are routed also depends on the forwarding policies configured for the listener.

Load Balancing Algorithms

Dedicated load balancers support four load balancing algorithms: weighted round robin, weighted least connections, source IP hash, and connection ID.

Shared load balancers support weighted round robin, weighted least connections, and source IP hash.

Weighted Round Robin

Figure 2 shows an example of how requests are distributed using the weighted round robin algorithm. Two backend servers are in the same AZ and have the same weight, and each server receives the same proportion of requests.

Figure 2 Traffic distribution using the weighted round robin algorithm

**Table 1** Weighted round robin
Description	Requests are routed to backend servers in sequence based on their weights. Backend servers with higher weights receive proportionately more requests, whereas equal-weighted servers receive the same number of requests.
When to Use	This algorithm is typically used for short connections, such as HTTP connections. Flexible load balancing: When you need more refined load balancing, you can set a weight for each backend server to specify the percentage of requests to each server. For example, you can set higher weights to backend servers with better performance so that they can process more requests. Dynamic load balancing: You can adjust the weight of each backend server in real time when the server performance or load fluctuates.
Disadvantages	You need to set a weight for each backend server. If you have a large number of backend servers or your services require frequent adjustments, setting weights would be time-consuming. If the weights are inappropriate, the requests processed by each server may be imbalanced. As a result, you may need to frequently adjust server weights.

Weighted Least Connections

Figure 3 shows an example of how requests are distributed using the weighted least connections algorithm. Two backend servers are in the same AZ and have the same weight, 100 connections have been established with backend server 01, and 50 connections have been established with backend server 02. New requests are preferentially routed to backend server 02.

Figure 3 Traffic distribution using the weighted least connections algorithm

**Table 2** Weighted least connections
Description	Requests are routed to the server with the lowest connections-to-weight ratio. In addition to the number of connections, each server is assigned a weight based on its capacity. Requests are routed to the server with the lowest connections-to-weight ratio.
When to Use	This algorithm is often used for persistent connections, such as connections to a database. Flexible load balancing: Load balancers distribute requests based on the number of established connections and the weight of each backend server and route requests to the server with the lowest connections-to-weight ratio. This helps prevent servers from being underloaded or overloaded. Dynamic load balancing: When the number of connections to and loads on backend servers change, you can use the weighted least connection algorithm to dynamically adjust the requests distributed to each server in real time. Stable load balancing: You can use this algorithm to reduce the peak loads on each backend server and improve service stability and reliability.
Disadvantages	Complex calculation: The weighted least connections algorithm needs to calculate and compare the number of connections established with each backend server in real time before selecting a server to route requests. Dependency on connections to backend servers: The algorithm routes requests based on the number of connections established with each backend server. If monitoring data is inaccurate or outdated, requests may not be distributed evenly across backend servers. The algorithm can only collect statistics on the connections between a given load balancer and a backend server, but cannot obtain the total number of connections to the backend server if it is associated with multiple load balancers. Too many loads on new servers: If existing backend servers have to handle a large number of requests, new requests will be routed to new backend servers. This may deteriorate new servers or even cause them to fail.

Source IP Hash

Figure 4 shows an example of how requests are distributed using the source IP hash algorithm. Two backend servers are in the same AZ and have the same weight. If backend server 01 has processed a request from IP address A, the load balancer will route new requests from IP address A to backend server 01.

Figure 4 Traffic distribution using the source IP hash algorithm

**Table 3** Source IP hash
Description	The source IP hash algorithm calculates the source IP address of each request and routes requests from the same IP address to the same backend server.
When to Use	This algorithm is often used for applications that need to maintain user sessions or state. Session persistence: Source IP hash ensures that requests with the same source IP address are distributed to the same backend server. Data consistency: Requests with the same hash value are distributed to the same backend server. Load balancing: In scenarios that have high requirements for load balancing, this algorithm can distribute requests to balance loads among servers.
Disadvantages	Imbalanced loads across servers: This algorithm tries its best to ensure request consistency when backend servers are added or removed. If the number of backend servers decreases, some requests may be redistributed, causing imbalanced loads across servers. Complex calculation: This algorithm calculates the hash values of requests based on hash factors. If servers are added or removed, some requests may be redistributed, making calculation more difficult.

Connection ID

Figure 5 shows an example of how requests are distributed using the connection ID algorithm. Two backend servers are in the same AZ and have the same weight. If backend server 01 has processed a request from client A, the load balancer will route new requests from client A to backend server 01.

Figure 5 Traffic distribution using the connection ID algorithm

**Table 4** Connection ID
Description	The connection ID algorithm calculates the QUIC connection ID and routes requests with the same hash value to the same backend server. A QUIC ID identifies a QUIC connection. This algorithm distributes requests by QUIC connection. You can use this algorithm to distribute requests only to QUIC backend server groups.
When to Use	This algorithm is typically used for QUIC requests. Session persistence: The connection ID algorithm ensures that requests with the same hash value are distributed to the same backend server. Data consistency: Requests with the same hash value are distributed to the same backend server. Load balancing: In scenarios that have high requirements for load balancing, this algorithm can distribute requests to balance loads among servers.
Disadvantages	Imbalanced loads across servers: This algorithm tries its best to ensure request consistency when backend servers are added or removed. If the number of backend servers decreases, some requests may be redistributed, causing imbalanced loads across servers. Complex calculation: This algorithm calculates the hash values of requests based on hash factors. If servers are added or removed, some requests may be redistributed, making calculation more difficult.

Factors Affecting Load Balancing

In addition to the load balancing algorithms, factors that affect load balancing include the connection type (persistent connection or short connection), session stickiness, and server performance and weights.

Assume that there are two backend servers with the same weight (not zero), the weighted least connections algorithm is selected, sticky sessions are not enabled, and 100 connections have been established with backend server 01, and 50 connections with backend server 02.

When client A wants to access backend server 01, the load balancer establishes a persistent connection with backend server 01 and continuously routes requests from client A to backend server 01 before the persistent connection is disconnected. When other clients access backend servers, the load balancer routes the requests to backend server 02 using the weighted least connects algorithm.

If backend servers are declared unhealthy or their weights are set to 0, the load balancer will not route any request to the backend servers.

For details about the load balancing algorithms, see Load Balancing Algorithms.

If requests are not evenly routed, troubleshoot the issue by performing the operations described in How Do I Check Whether Traffic Is Evenly Distributed?

Previous topic: ELB Advantages

Next topic: Application Scenarios