Global Accelerator:Use GA traffic distribution to switch traffic across regions

How it works

Global Accelerator (GA) provides two independent controls for traffic:

• Traffic distribution ratio — the percentage of traffic directed to an endpoint group. Set at the endpoint group level. Valid values: 0–100. The default is 100.

• Endpoint weight — how traffic is distributed among endpoints within a single endpoint group. Valid values: 0–255. The default is 255.

When multiple endpoint groups are under one listener, GA also applies scheduling priority: the endpoint group with lower network latency (determined by proximity to the access point) gets higher priority and receives traffic first.

The final traffic distribution depends on both the ratio and the scheduling priority of each endpoint group.

Key behaviors:

• The ratio controls what percentage of traffic is forwarded to an endpoint group. Setting the ratio to 0% stops all traffic to that endpoint group; setting it to 100% sends all eligible traffic to it. When you reduce the ratio of the higher-priority endpoint group, traffic is redistributed to other endpoint groups according to their ratios and scheduling priorities.

• If health checks are enabled and a higher-priority endpoint group fails, all traffic moves to the next available endpoint group — the configured ratio is ignored until recovery.

• If a forwarding rule exists, traffic is scheduled among the endpoint groups associated with the matched forwarding rule.

For the subscription billing method, only TCP and UDP listeners support traffic distribution. For the pay-as-you-go billing method, all listener types support traffic distribution.

Use cases

This tutorial walks through a regional migration scenario.

Scenario

A company runs a service on servers in the China (Beijing) and China (Shanghai) regions. The TCP protocol is used on port 80, and clients are in the China (Beijing) region.

By default, GA sends all client requests to the endpoint group in China (Beijing). China (Shanghai) acts as the secondary endpoint group. The company wants to migrate traffic to China (Shanghai) while keeping the service available throughout the switchover.

The migration uses three stages:

Prerequisites

Before you begin, ensure that you have:

• Two servers deployed — one in China (Beijing) and one in China (Shanghai) — both running on TCP, port 80

• An Alibaba Cloud account with permission to create GA instances

• (Optional) Activated the pay-as-you-go GA service at Activate the Service if this is your first time using pay-as-you-go GA instances

This tutorial uses a pay-as-you-go standard GA instance. Pay-as-you-go instances use the pay-by-data-transfer metering method. No basic bandwidth plan is required — data transfer billing is managed by Cloud Data Transfer (CDT). For billing details, see Pay-by-data-transfer and Billing of pay-as-you-go GA instances.

Step 1: Deploy and verify servers

The servers in this example run Alibaba Cloud Linux 3.2104 (64-bit). Commands may vary depending on your operating system.

1. Deploy servers in China (Beijing) and China (Shanghai). Open port 80 on TCP for both servers.

2. From a client in China (Beijing), run the following command to confirm that both servers are reachable:

   curl <Origin server IP address>

   Each server returns its region information. Confirm that both servers respond correctly before continuing. Figure 1. Response from the China (Beijing) server Figure 2. Response from the China (Shanghai) server

Step 2: Create a GA instance

1. Log on to the GA console.

2. On the Instances page, click Create Standard Pay-as-you-go Instance.

3. In the Basic Instance Configuration step, set the following parameters and click Next.

   Parameter	Description
   GA Instance Name	Enter a name for the GA instance.
   Instance Billing Method	Pay-As-You-Go is selected by default. You are charged instance fees, Capacity Unit (CU) fees, and data transfer fees.
   Resource Group	Select the resource group for this instance. The resource group must be created by the current Alibaba Cloud account in Resource Management. See Create a resource group.

Step 3: Configure an acceleration area

Specify the acceleration area and allocate bandwidth.

In the Configure Acceleration Area step, set the following parameters and click Next.

Step 4: Configure a listener

A listener receives connection requests and routes them to endpoint groups based on the port and protocol you specify.

In the Configure listeners step, set the following parameters and click Next. For parameters not listed here, keep the defaults. For full parameter descriptions, see TCP or UDP listener.

Step 5: Configure endpoint groups

Step 6: Test traffic distribution

Use the following command to simulate client requests and measure how traffic is distributed between regions:

echo > curl.txt; for ((i=0;i<<Number of requests>;i++)); do curl -s <Accelerated IP address> >> curl.txt; done; beijing_count=`grep Beijing curl.txt | wc -l`;echo "Beijing count: ${beijing_count}";shanghai_count=`grep Shanghai curl.txt | wc -l`;echo "shanghai count: ${shanghai_count}";

Replace the placeholders before running:

The command counts responses from Beijing (Beijing count) and Shanghai (shanghai count).

Stage 1: Baseline (Beijing ratio = 100%)

Run the command from a client in China (Beijing) with 100 requests.

All 100 requests are handled by the China (Beijing) endpoint group. Because Beijing has higher scheduling priority for clients in that region and the ratio is 100%, no traffic reaches Shanghai.

What's next

• Enable and manage health checks — configure health checks so that GA automatically reroutes traffic if an endpoint group becomes unavailable.

• Distribute traffic across endpoint groups in different scenarios — learn the principles behind traffic distribution across multiple endpoint groups.

• Use GA to accelerate applications deployed across regions and implement disaster recovery — combine traffic distribution and health checks for high-availability disaster recovery.