After you create a standard Global Accelerator (GA) instance, you need to configure listeners for it. A listener checks connection requests from clients and distributes traffic to endpoints based on the forwarding method defined by its routing type.
You can create up to 50 listeners for each GA instance.
Listener protocols
Listeners support four protocols: TCP, UDP, HTTP, and HTTPS. Choose a protocol based on your application requirements.
| Protocol | Description | Scenarios |
|---|---|---|
| TCP | A connection-oriented protocol that provides high reliability. A connection must be established before data can be sent or received. Supports session persistence based on source IP addresses. Source IP addresses are visible at the network layer. Data is transmitted at a slow rate. | Suitable for scenarios that require high reliability and data accuracy but can tolerate slower transmission speeds, such as file transfers, email, and remote logons. Also suitable for web applications that do not have special requirements. |
| UDP | A connectionless protocol with low reliability. No three-way handshake is required before data is sent, and the protocol does not provide error recovery or data retransmission. Data is transmitted at a high rate. | Suitable for scenarios that prioritize real-time performance over reliability, such as video chats and real-time financial market data pushes. |
| HTTP | A connection-oriented protocol that provides high reliability. A connection must be established before data can be sent or received. Data is transmitted at a high rate in plaintext. | Suitable for accelerating access to HTTP websites and improving client access speed. Also suitable for accelerating access to HTTP websites that use multiple domain names or paths. |
| HTTPS | A connection-oriented protocol that provides high reliability. A connection must be established before data can be sent or received. Uses an SSL certificate to ensure high data reliability. Data is encrypted during transmission. You can configure the maximum HTTP protocol version to enable HTTP/3 access, which further improves the stability, security, and efficiency of data transmission. Note For more information about SSL certificates, see What is Digital Certificate Manager. | Suitable for accelerating access to HTTPS websites and improving the speed and security of client access. Also suitable for accelerating access to HTTPS websites that use multiple domain names or paths. |
Listener routing types
Listeners support two routing types: smart routing and custom routing. The following table summarizes when to use each type.
| Smart routing | Custom routing | |
|---|---|---|
| Best for | Applications that require fine-grained control over global traffic distribution | Applications that require precise mapping of clients to specific backend services |
| How it works | GA automatically selects a nearby, healthy endpoint group based on latency factors | GA uses a port mapping table to route traffic to specific IP addresses and ports |
| Supported protocols | TCP, UDP, HTTP, HTTPS | TCP, UDP only |
| Example use cases | Blue-green deployments, A/B testing, multi-region deployments, multi-region disaster recovery | Multiplayer games, video conferencing, virtual classrooms |
| Availability | Default | Invitational preview |
The custom routing listener feature is available for invitational preview. To use this feature, contact your account manager. After your request is approved, you can select the routing type for your listeners.
A single GA instance cannot simultaneously support listeners of different routing types. The routing type cannot be changed after the listener is configured. For more information about the functional differences between the two routing types, see Comparison between smart routing and custom routing listeners.
Smart routing listeners
Smart routing is the default routing type for a standard listener. With smart routing, GA automatically selects a nearby and healthy endpoint group based on latency factors such as geographic location and network link conditions, then forwards client requests to the optimal endpoint.
This routing type is suitable for applications that require fine-grained control over global traffic. You can configure traffic weights for endpoint groups and endpoints to support scenarios such as blue-green deployments, A/B testing, multi-region deployments, phasing out or upgrading services by region, and multi-region disaster recovery. For more information about application scenarios and examples, see Traffic distribution principles and scenarios for multiple endpoint groups and Use GA to smoothly switch traffic across regions.
Smart routing listeners support the TCP, UDP, HTTP, and HTTPS protocols.
Custom routing listeners
With custom routing, GA generates a port mapping table based on the configured listener port range, the destination port range of the endpoint group, and the IP addresses of the endpoints (vSwitches). The listener uses this port mapping table to route client traffic to specific IP addresses and ports on a vSwitch.
This routing type is suitable for applications that require precise mapping of clients to backend services, such as assigning multiple users to specific servers. Example use cases include multiplayer games, video conferencing, and virtual classrooms. For more information about how custom routing listeners work and for usage examples, see How custom routing listeners work.
Custom routing listeners support only the TCP and UDP protocols. You must specify the protocols when you configure the endpoint group. You can specify one or more protocols for each configured destination port range of an endpoint group -- TCP, UDP, or both.
Listener ports
A listener port receives requests from clients and forwards them to endpoints. Port configuration rules differ between smart routing and custom routing listeners.
Modifying listener ports or the protocol after configuration may cause traffic interruptions. Perform these modifications during off-peak hours or after you switch traffic to another GA instance.
Smart routing listener ports
For smart routing listeners on the same GA instance:
The ports of TCP, HTTP, and HTTPS listeners cannot be the same.
The port of a UDP listener cannot be the same as the port of an HTTPS listener for which HTTP/3 is enabled.
The following table describes the configurable port range and port quota for each listener protocol.
| Listener protocol | Configurable listener port range | Listener port quota |
|---|---|---|
| TCP | 1 to 65499 | 30 |
| UDP | 1 to 65499 | 30 |
| HTTP | 1 to 65499 | 1 |
| HTTPS | 1 to 65499 | 1 |
For TCP and UDP listeners, separate multiple ports with commas (,). Example: 80,90,8080. Use a hyphen (-) to specify a port range. Example: 80-83 specifies ports 80, 81, 82, and 83.
Additional limits apply to TCP and UDP listeners depending on the billing method of your GA instance:
Pay-as-you-go mode
You can configure multiple port ranges that contain consecutive ports for a single listener. The configurable port range is 1 to 65499. The port ranges cannot overlap.
The number of ports you can add to a single listener is controlled by the gaplus_quota_port_per_listener quota. Each configured port or port range consumes one quota unit.
The total number of ports and port ranges you can add across all listeners on a single GA instance is controlled by the gaplus_quota_listener_per_accelerator quota. The combined count of individual ports and port ranges across all listeners cannot exceed this quota.
The default value of the gaplus_quota_listener_per_accelerator quota is 50. To request a quota increase, contact your business manager.
Example: Suppose you configure a TCP listener with the ports 1,3-350,400-800. This configuration contains three entries (one individual port and two port ranges), so it consumes 3 units of the gaplus_quota_port_per_listener quota. You can then add up to 27 more ports or port ranges to this TCP listener. If this TCP listener reaches its quota of 30 entries, you can still add up to 20 ports or port ranges to other TCP or UDP listeners on the same GA instance.
Subscription mode
The number of ports you can add to a single listener is controlled by the gaplus_quota_port_per_listener quota. Each individual port consumes one quota unit. Example: If you configure a TCP listener with ports
1,3-5, four ports are specified (1, 3, 4, and 5), so 4 units of the gaplus_quota_port_per_listener quota are consumed. You can add up to 26 more ports to this listener.By default, you can configure a port range of more than 300 consecutive ports for a single listener. A listener with more than 300 consecutive ports is called a massive-port listener. The following rules apply: Example: Suppose you need to configure the following listeners for a GA instance: TCP 1-400, TCP 443, UDP 200-210, and UDP 230-240. In this case, TCP 1-400 is a massive-port listener because it contains more than 300 consecutive ports.
If you cannot configure a port range of more than 300 consecutive ports for a TCP or UDP listener, the GA instance version may not support this feature. Contact your business manager to upgrade the instance.
You must configure more than 300 ports. You can configure up to 65,499 ports.
Each subscription GA instance supports only one massive-port listener.
A massive-port listener consumes one unit of the gaplus_quota_port_per_listener quota.
Custom routing listener ports
The following rules apply to custom routing listener ports:
The configurable port range is 1 to 65499. Ports 25, 250, 4789, and 4790 are reserved by the system. The system automatically ignores reserved ports when it generates a port mapping table. Pay-as-you-go GA instances do not support port 6081.
The listener port range you configure determines the number of port-and-IP-address combinations available in the associated endpoint group. The number of listener ports (excluding system-reserved ports) must satisfy the following formula:
Number of listener ports >= Total number of ports in the endpoint group x Total number of IP addresses of all vSwitches in the endpoint groupWe recommend that you configure a large port range for the listener. Example: If the destination port range of your endpoint group is81-85(5 ports) and the total number of IP addresses across all vSwitch endpoints is 16, the listener port range must contain at least 80 ports (5 x 16 = 80). A listener port range of101-180would work, but101-179would not. If the listener port range is too small, the listener cannot be created.After you configure the listener ports, you cannot remove ports that have existing mappings when you modify the port range. Example: If the original listener port range is
100-10000and a mapping exists between port199and destination port80of an endpoint, you can expand the range to20-10000but you cannot narrow it to200-10000.Listener ports of different listeners on the same GA instance cannot overlap.