A basic Global Accelerator (GA) instance provides a one-to-one acceleration network from an acceleration region to the region where the origin resides, and supports Layer 3 (IP) network acceleration.
When you use basic GA, the backend service must be deployed on Alibaba Cloud. Clients can connect only through an accelerated IP, and the instance is not assigned an acceleration CNAME. If you need Layer 4 or Layer 7 application acceleration, or you need to resolve a business domain name to the acceleration network, use standard GA.
Scenario
A company deploys an enterprise application service on Alibaba Cloud ECS in the US (Silicon Valley) region. Employees at its China (Hong Kong) office must access the service over the public network. Because the public network link is unstable, they frequently experience latency, jitter, and packet loss.
After Global Accelerator is configured, traffic from the China (Hong Kong) office enters the Alibaba Cloud acceleration network through an accelerated IP at the nearby China (Hong Kong) access point, and is then forwarded to the endpoint in the US (Silicon Valley) region. This improves access speed and stability.
1. Create a basic instance
Global Accelerator supports two billing methods: Pay-as-you-go and Subscription. You can select a billing method only when you purchase a GA instance. After an instance is created, the billing method cannot be changed.
Pay-as-you-go: Recommended when resource usage is temporary or bursty, traffic peaks fluctuate widely, or resource and traffic usage cannot be estimated. The first time you use a pay-as-you-go GA instance, go to the service activation page and follow the instructions to activate the pay-as-you-go GA service.
Subscription: Recommended when the usage period can be estimated, traffic peaks are stable, and you need the resources long term.
Instance billing method | Bandwidth billing method | Billing description |
Pay-as-you-go | Pay-by-data-transfer | No bandwidth plan is required. Traffic fees incurred for access to the acceleration network are settled and billed uniformly by Cloud Data Transfer (CDT). |
Subscription | Pay-by-bandwidth | A subscription instance must purchase and bind a basic bandwidth plan to settle bandwidth fees. If your workloads involve acceleration between the Chinese mainland and Hong Kong (China), Macao (China), Taiwan (China), or other countries and regions, you must additionally purchase a cross-border acceleration bandwidth plan and use it together with the basic bandwidth plan. |
Console
Go to the GA console - Basic Instances page and click Create Basic Instance.
Select the Instance Billing Method that you need. The Bandwidth Billing Method changes accordingly. This topic uses Pay-As-You-Go.
If you select Subscription, you must also set the Subscription Duration. After a subscription instance is created, go to the Bandwidth Plans page and click Purchase Basic Bandwidth Plan. After the purchase, click Bind Instance in the Actions column of the target basic bandwidth plan, and then Select Instance. After the bandwidth plan is bound, its status is Available.
API
Call CreateBasicAccelerator to create a basic instance, and select the ChargeType that you need.
After a subscription instance is created, call CreateBandwidthPackage to create a bandwidth plan, and call BandwidthPackageAddAccelerator to bind the bandwidth plan to the instance.
2. Configure an acceleration area
An acceleration area specifies the regions where the clients that need accelerated access to the backend service are located. We recommend that you select the acceleration region where your clients are located. If that region is not available, select the acceleration region closest to your clients.
After you add an acceleration area and an acceleration region, you must allocate an accelerated IP to the acceleration region and bind it to an endpoint. Only then can clients enter the Alibaba Cloud acceleration network from the acceleration region by accessing the accelerated IP.
Acceleration area: A basic instance supports only one acceleration area, and only one acceleration region can be added to that area.
Bandwidth: For a pay-as-you-go instance, you can customize the bandwidth. For a subscription instance, the bandwidth is determined by the basic bandwidth plan bound to the instance.
If the bandwidth is set too low, rate limiting may cause traffic to be dropped. Plan the bandwidth peak reasonably to match your business requirements.
IP protocol: Basic instances support only IPv4.
Public network quality type: BGP (Multi-ISP) (default, available in all acceleration regions) and BGP (Premium) (available only in China (Hong Kong) and Japan (Tokyo), which optimizes traffic from overseas back to the Chinese mainland).
Only pay-as-you-go instances require you to select a public network quality type.
Accelerated IP: An accelerated IP provides service only after it is bound to an endpoint. An accelerated IP that is not bound to an endpoint is in the Idle state.
By default, a single instance supports up to 20 accelerated IPs, of which up to 10 can be in the Idle state. The two quotas are controlled by
gaplus_quota_basic_gaip_limitandgaplus_quota_basic_gaip_idle_limitrespectively.After the quota for idle accelerated IPs is reached, you can neither add accelerated IPs nor unbind existing ones.
Basic instances created before August 1, 2023 use a different specification from the one described in this topic: you do not need to add and bind accelerated IPs separately, only one endpoint group and one endpoint are supported, and NLB is not supported as a backend service. To use the current specification, contact your account manager to upgrade the instance.
Console
On the Basic GA Instance page, click the ID of the target instance.
Select the Acceleration Areas tab, click Add Acceleration Area, and configure the Regions, Bandwidth, Internet Protocol, and ISP Line Type.
In this topic, the acceleration region is set to China (Hong Kong), and ISP Line Type is set to BGP (Multi-ISP).
After you add the acceleration area, click Add Accelerated IP Address.
If an endpoint group and an endpoint are already configured, you can select an Endpoint and click OK.
If no endpoint is configured yet, click Cancel. The allocated accelerated IP is in the Idle state.
After the acceleration area is added, a pay-as-you-go instance can adjust the bandwidth by using Edit Bandwidth, whereas a subscription instance must adjust it through the bound bandwidth plan.
API
Call CreateBasicIpSet to add an acceleration area. Specify
AcceleratorId, and configureAccelerateRegionId,Bandwidth, andIspType(pay-as-you-go instances only).After the acceleration region is added, to adjust the bandwidth, call UpdateBasicIpSet for a pay-as-you-go instance, or UpdateBandwidthPackage for a subscription instance.
3. Configure an endpoint group
Each endpoint group is associated with a specific region. By specifying the region to which traffic is distributed, a basic instance distributes traffic to the endpoints in the endpoint group. For a basic instance, you must bind accelerated IPs to endpoints. The binding creates a mapping, and GA distributes traffic to the corresponding backend service based on the mapping that you specify.
Endpoint group region: The region where the backend service is deployed. If that region is not in the list of supported regions, select the region closest to the backend. A basic instance supports only one endpoint group.
Endpoint configuration: Select the backend service type and the backend service instance of the endpoint. A basic instance can connect to the backend service only over a private network. The supported types are ECS, ENI (secondary ENI only), CLB (private), and NLB (private).
You cannot select a backend service instance that has an EIP or a static public IP assigned. If one is assigned, you must first unbind the EIP. A static public IP must be converted to an EIP before it can be unbound.
For a pay-as-you-go instance, if your configuration involves acceleration between regions in the Chinese mainland and China (Hong Kong), China (Macao), China (Taiwan), or other countries and regions, make sure that the current account has completed enterprise real-name registration.
You cannot delete an endpoint group that contains an endpoint whose IP Status is Bound. When the IP Status of all endpoints is Idle, you can delete the endpoint group directly, and the system deletes all endpoints in the endpoint group at the same time. After the endpoint group is deleted, GA no longer provides acceleration for it.
Console
On the Basic GA Instance page, click the ID of the target instance.
Select the Endpoint Group tab, click Add Endpoint Group, and configure the Endpoint Group Name, Region, and Endpoint.
You must select the Private IP Address of the backend service instance, and then either select an idle accelerated IP or add an accelerated IP to bind to the endpoint.
If you do not select one here, you can, after the endpoint group is created, select Bind Accelerated IP Address on the Endpoint Group tab, or select Associate Endpoint on the Acceleration Areas tab.
In this topic, the Region is set to US (Silicon Valley), the backend service is set to ECS, and the accelerated IP added in Step 2 is bound to it.
Required in specific cases: For a pay-as-you-go instance, if your configuration involves acceleration between regions in the Chinese mainland and China (Hong Kong), China (Macao), China (Taiwan), or other countries and regions, in the Cross-border Acceleration Settings dialog box that appears, read the Compliance Commitments Regarding Cross-border Data Transfers and select Agree to the Preceding Compliance Agreement.
After endpoints are added, you can manage endpoint groups and endpoints on the Endpoint Group tab of the instance details page.
API
Call CreateBasicEndpointGroup to create an endpoint group. Specify
AcceleratorId, and configureEndpointGroupRegion,EndpointType, andEndpointAddress.Call CreateBasicAccelerateIpEndpointRelation to bind an accelerated IP to an endpoint.
4. Test the acceleration performance
After the configuration is complete, run ping <GA accelerated IP> on the client to check packet latency and packet loss and verify the acceleration performance.
To compare acceleration performance, we recommend that you run ping <public IP of the backend service> before deployment to record baseline data.