Global Accelerator:Key concepts

Acceleration area

An acceleration area is a geographic area where you want to accelerate access to your applications. Each acceleration area contains one or more regions. You can select the region where your clients are located, or a nearby region, as the acceleration region. For more information, see Acceleration areas and regions supported by standard GA instances and Acceleration areas and regions supported by basic GA instances.

After a client connects through the accelerated IP address or CNAME assigned by GA, traffic is always forwarded through the selected acceleration node. GA does not dynamically evaluate the latency difference between the direct and accelerated paths, nor does it automatically fall back to a direct connection.

• The number of acceleration areas and regions you can add to a standard GA instance depends on its billing method:

  • pay-as-you-go: The limit is based on the total number of acceleration areas and regions supported by GA.

  • subscription: The limit is determined by the instance specification.

• A basic GA instance supports only one acceleration area and one region.

Accelerated IP

Clients use an accelerated IP address to connect to the Alibaba Cloud acceleration network through the nearest region.

• A standard GA instance supports the following two types of accelerated IP addresses:

  • EIP: This type uses a custom nearby access mode. You can select specific acceleration areas and regions based on your business needs. Each acceleration region provides a separate EIP.

  • Anycast EIP: This type uses an automatic nearby access mode. You do not need to configure acceleration areas. GA provides a single, unified IP entry point across multiple regions worldwide.

• A basic GA instance uses an EIP as its accelerated IP address by default.

  You can manually add multiple accelerated IP addresses to the acceleration region of a basic GA instance. You can then create mappings between accelerated IP addresses and endpoints to route traffic to the appropriate backend services.

CNAME

A CNAME is a canonical name record assigned to a GA instance.

To accelerate traffic for a custom domain, create a DNS record that maps your domain name to the GA instance's CNAME so that requests are routed through GA.

• When you create a standard GA instance, the system automatically assigns a CNAME to it.

• CNAME configuration is not required for a basic GA instance.

Listener

A listener monitors connection requests and routes inbound client traffic based on the port and routing type that you specify.

• A standard GA instance supports the following two routing types:

  • intelligent routing: This is the default routing type for a standard listener. It automatically routes traffic to the nearest or healthiest endpoint group based on factors like latency, ensuring that client requests reach the optimal endpoint.

  • custom routing: This routing type deterministically routes traffic to a specific IP address and port within a designated vSwitch. The routing decision is based on the mapping between the listener's port and the CIDR block of the backend vSwitch.

  You can select the protocol and routing type based on your use case.

• Listener configuration is not required for a basic GA instance.

Endpoint group

An endpoint group receives traffic based on the forwarding method defined by the listener's routing type. Each endpoint group is associated with a specific region and contains one or more endpoints.

• For a standard GA instance, the types of endpoint groups you can create depend on the listener's routing type and protocol:

  • intelligent routing: You can create multiple default endpoint groups and virtual endpoint groups.

    • default endpoint group: The endpoint group you configure when creating a listener is the default endpoint group.

      By default, a default endpoint group is associated with the default forwarding policy. You can also manually associate it with a custom forwarding policy.

    • virtual endpoint group: After creating a listener, you can manually create a virtual endpoint group on the Endpoint Group page.

      A virtual endpoint group can only be associated with a custom forwarding policy.

  • custom routing: This routing type does not differentiate between endpoint group types.

• For a basic GA instance, you do not need to configure a listener. You can directly add an endpoint group for a specific region to the instance.

  A basic GA instance supports only one endpoint group.

Endpoint

An endpoint is the destination server or resource that processes client requests.

• For a standard GA instance, the supported backend service types for an endpoint vary based on where the service is deployed.

  Deployment location	Backend service type
  On Alibaba Cloud	ECS ALB NLB CLB OSS ENI custom private IP address Alibaba Cloud Public IP vSwitch This endpoint type is supported only by custom routing listeners.
  Outside Alibaba Cloud	custom public IP address Custom Domain Name

  For more information, see Endpoints for standard GA instances.

• A basic GA instance supports the following endpoint types:

  • ENI

  • ECS

  • CLB

  • NLB

  For more information, see Endpoints for basic GA instances.

Endpoint egress IP

GA translates the client's source IP address to an endpoint egress IP address before forwarding traffic to the endpoint. This translation applies only when GA connects to the endpoint over the public internet.

Endpoint egress IP addresses are dedicated to a single GA instance and not shared across users.

• After configuring endpoints for a standard GA instance, the system allocates at least four endpoint egress IP addresses for each endpoint. The actual number may vary based on the instance specification or traffic volume. Refer to the console for the exact number.

• Endpoint egress IP addresses are not used for basic GA instances.