Server Load Balancer:Disable cross-zone load balancing to reduce access latency

Limitations

Before you configure this feature, check the following constraints:

• Cross-zone load balancing can be disabled only on server groups associated with standard or WAF-enabled ALB instances. Basic ALB instances are not supported.

• Session Persistence cannot be enabled when cross-zone load balancing is disabled.

• For server groups of the IP type, cross-zone load balancing cannot be disabled if Remote IP is enabled.

• This setting is not available for server groups of the Function Compute type.

How it works

When cross-zone load balancing is enabled (the default), each ALB node can forward requests to any backend server across all zones within the specified region. When cross-zone load balancing is disabled on a server group, each ALB node routes traffic only to backend servers in its own zone. Requests that arrive at Zone H go exclusively to Zone H's backend servers, and Zone K traffic stays within Zone K.

Use cases

Disable cross-zone load balancing when your workload requires consistently low latency:

• Financial trading: Quantitative trading strategies are sensitive to millisecond-level delays. Keeping requests within a single zone reduces the risk of transaction timeouts caused by cross-zone network latency.

• Real-time communications: Video conferencing and instant messaging require low, stable latency for a smooth user experience.

• IoT data collection: Real-time sensor data pipelines need fast, predictable response times to maintain data accuracy and system efficiency.

Prerequisites

Before you begin, ensure that you have:

• A standard ALB instance deployed across multiple zones. In this example, the instance is deployed in Zone H and Zone K in the China (Hangzhou) region. See Create and manage an ALB instance

• A server group with backend servers in each zone. In this example, ECS01 is in Zone H and ECS02 is in Zone K, each with a web service deployed. See Create and manage server groups

  Sample commands to deploy a test service on ECS01 and ECS02

  ECS01 (Zone H):

  yum install -y nginx
  systemctl start nginx.service
  cd /usr/share/nginx/html/
  echo "Hello World ! This is ECS01." > index.html

  ECS02 (Zone K):

  yum install -y nginx
  systemctl start nginx.service
  cd /usr/share/nginx/html/
  echo "Hello World ! This is ECS02." > index.html

• Listeners configured on the ALB instance, with the server group associated. See Add an HTTP listener, Add an HTTPS listener, and Add a QUIC listener

• A registered domain name with an ICP (Internet Content Provider) filing, and a CNAME record pointing to the ALB instance's domain name. See Register a generic domain name, ICP filing process, and Configure a CNAME record for an ALB instance

Disable cross-zone load balancing

These steps show how to update an existing server group. If you are creating a new server group, you can disable cross-zone load balancing during creation.

1. Log on to the ALB console.

2. On the Server Groups page, click the ID of the server group you want to update.

3. On the Details page, in the Basic Information section, click Modify Basic Information.

4. In the Modify Basic Information dialog box, click Advanced Settings, turn off Cross-zone Load Balancing, then click Save.

Verify that cross-zone load balancing is disabled

1. On the Instance Details page of your ALB instance, open the Zone tab and copy the Elastic IP Address (EIP) for each zone.

2. Open a browser, navigate to http://<EIP_of_Zone_H>, and refresh several times. All responses should come from ECS01 in Zone H only.

   

3. Navigate to http://<EIP_of_Zone_K> and refresh several times. All responses should come from ECS02 in Zone K only.

   

Test the network latency

1. Create an internet-facing ECS instance in the Germany (Frankfurt) region to act as the test client. If you already have a suitable test machine, skip this step.

2. Log on to the test client and run:

   curl http://www.example.com -s -w "time_connect: %{time_connect}\ntime_starttransfer: %{time_starttransfer}\ntime_total: %{time_total}\n"

   The three metrics in the output:

   Metric	Description
   time_connect	Time to establish a TCP connection, in seconds
   time_starttransfer	Time from sending the request to receiving the first response byte, in seconds
   time_total	Total time from sending the request to receiving the last response byte, in seconds

3. Compare the results before and after disabling cross-zone load balancing.

   With cross-zone load balancing enabled:

   

   With cross-zone load balancing disabled:

   

   After disabling cross-zone load balancing, connection time, data transfer time, and total time all decrease.

FAQ

What's next

• For ALB regions and zones, see Regions and zones.

• For general ALB troubleshooting, see FAQ about ALB.

• For health check troubleshooting, see Troubleshoot ALB health check issues.

• API reference: CreateServerGroup, UpdateServerGroupServersAttribute, ListServerGroups, and GetListenerHealthStatus.