Selecting a Load Balancer That Fits Your Service Requirements

Scenarios

ELB provides dedicated and shared load balancers for you to choose based on your service size and type.

Dedicated load balancers provide fixed and elastic specifications. You can estimate the price based on your service size by referring to Elastic Load Balance Price Calculator.

Elastic specifications work well for fluctuating traffic, and you will be billed for how many LCUs you use.
Fixed specifications are suitable for stable traffic, and you will be billed for each specification you select.

This practice helps you select a load balancer that suits your service scenario.

Load Balancer Type Comparison

ELB provides dedicated and shared load balancers.

Dedicated load balancers work better than shared load balancers. Dedicated load balancers support a wider range of protocols and can route Layer 7 requests based on advanced forwarding policies you configure.

For details, see Differences Between Dedicated and Shared Load Balancers.

**Table 1** Load balancer types
Item	Dedicated Load Balancer	Shared Load Balancer
Deployment mode	A dedicated load balancer gets dedicated resources. Its performance is not affected by the loads on other load balancers. In addition, there is a wide range of specifications available for you to choose from.	They are deployed in clusters and share resources with other instances. They support guaranteed performance.
Specifications	Elastic specifications: You are billed for how long a load balancer is running and the number of LCUs you use. Fixed specifications: Multiple specifications are available for you to select to best meet your needs.	N/A
Performance	A dedicated load balancer in an AZ can establish up to 20 million concurrent connections. If you select more than one AZ when creating a dedicated load balancer, the number of concurrent connections and new connections will be multiplied by the number of AZs. For details, see Specifications of Dedicated Load Balancers.	If guaranteed performance is enabled, each shared load balancer can handle up to 50,000 concurrent connections, 5,000 new connections per second, and 5,000 queries per second. The shared load balancer may not process extra requests if the guaranteed connection limit is exceeded.
AZ	You can select one or more AZs as needed.	N/A
Billed item	Fixed specifications: billed by the LCUs based on the specifications you select. Elastic specifications: billed by how many LCUs you use and how long a load balancer is retained in your account.	You are billed for how long you use a load balancer if guaranteed performance is enabled.

Key Feature Comparison

Compared with shared load balancers, dedicated load balancers support a wider range of protocols and can route Layer 7 requests based on advanced forwarding policies you configure.

For details, see Feature Comparison Details.

**Table 2** Feature comparison
Item	Dedicated Load Balancer	Shared Load Balancer
Capabilities	Powerful capabilities to process Layer 4 and Layer 7 requests, advanced forwarding policies, and multiple protocols	Basic capabilities to process Layer 4 and Layer 7 requests
Application scenarios	Heavy-traffic and highly concurrent services, such as large websites, cloud-native applications, IoV, and multi-AZ disaster recovery applications	Services with low traffic
Frontend protocols	TCP, UDP, HTTP, HTTPS, QUIC, and TLS	TCP, UDP, HTTP, and HTTPS
Backend protocols	TCP, UDP, HTTP, HTTPS, QUIC, TLS, and gRPC	TCP, UDP, and HTTP
Forwarding capabilities	Provides powerful Layer 4 and Layer 7 processing capabilities to forward requests based on the following: Forwarding rules: domain name, path, HTTP request method, HTTP header, query string, and CIDR block Actions: forward to a backend server group, redirect to another listener, redirect to another URL, rewrite, and return a specific response body	Provides basic Layer 4 and Layer 7 processing capabilities to forward requests based on the following: Forwarding rules: domain name and path Actions: forward to a backend server group and redirect to another listener
Key features of backend server groups	A backend server group can be associated with multiple load balancers and listeners. Forwarding mode: load balancing and active/standby forwarding	A backend server group can be associated with only one listener. Forwarding mode: Load balancing
Backend server	ECSs and supplementary network interfaces in the same VPC as the load balancer IP addresses of servers in different VPCs from the load balancer and on-premises data centers	Cloud servers in the same VPC as the load balancer

Differences Between Application and Network Load Balancing

You can select application load balancing, or network load balancing, or both as required.

Application load balancing (HTTP/HTTPS): supports HTTP, QUIC, and HTTPS. This option is a great fit for workloads that require high performance at Layer 7, such as real-time audio and video, interactive livestreaming, and game applications.
Networking load balancing (TCP/UDP/TLS): supports TCP, TLS, and UDP. This works well for heavy-traffic and high-concurrency workloads at Layer 4, such as file transfer, instant messaging, and online video applications.

**Table 3** Protocols supported by ELB
Type	Protocol	Description	Application Scenario
Network listeners	TCP	Source IP address–based sticky sessions Fast data transfer	Scenarios that require high reliability and data accuracy, such as file transfer, email, and remote login Web applications that do not need to handle a large number of concurrent requests and do not require high performance
Network listeners	UDP	Relatively low reliability Fast data transfer	Scenarios that require quick response, such as video chat, gaming, and real-time financial news
Network listeners	TLS	An extension of HTTP for encrypted data transmission that can prevent unauthorized access Unidirectional/Bidirectional authentication	Scenarios that require ultra-high performance and large-scale TLS offloading
Application listeners	HTTP	Cookie-based sticky sessions X-Forward-For request header	Applications that require content identification, for example, web applications and mobile games
Application listeners	HTTPS	An extension of HTTP for encrypted data transmission that can prevent unauthorized access Encryption and decryption performed on load balancers Multiple versions of encryption protocols and cipher suites	Workloads that require encrypted transmission, such as e-commerce and financial services
Application listeners	QUIC	UDP-based low-latency internet transport layer protocol Multiplexing without head-of-line blocking Improved congestion control	Applications with a poor network environment and whose users have to switch between networks