Global Distributed Cache is a self-developed active-active database system by Alibaba Cloud that links multiple Tair (Enterprise Edition) child instances across regions into a single logically distributed instance. All child instances can read and write simultaneously while staying in real-time sync — making it suited for active geo-redundancy, disaster recovery, and cross-region data synchronization.
This page shows you how to create a distributed instance by either converting an existing Tair memory-optimized instance or purchasing a new one as the first child instance.
How it works
Global Distributed Cache networks child instances — each deployed in a separate region — through data synchronization channels. When a write occurs on any child instance, it propagates to all others in real time.
| Use case | Description |
|---|---|
| Active geo-redundancy | Multiple regions serve traffic simultaneously, reducing latency for globally distributed users. |
| Disaster recovery | If one region becomes unavailable, other child instances continue serving traffic without manual failover. |
| Cross-region data synchronization | Data written in one region is automatically replicated to all participating instances. |
Prerequisites
Before you begin, make sure you have:
An Alibaba Cloud account with access to the ApsaraDB for Redis console
A Tair memory-optimized instance to convert, or a plan to purchase a new one as the first child instance
(For converting an existing instance) A standard connection address (for example,
r-hp3bpn39cs1vu****.redis.hangzhou.rds.aliyuncs.com) — direct connection addresses are not supported for conversion(For converting an existing instance) A reconnection mechanism in your application
The Global Distributed Cache feature itself is free. You are charged only for the specifications of child instances, using the same billing standard as standard instances. For details, see Billing items.
Instance limits
The first child instance can be converted from an existing Tair memory-optimized instance or newly purchased.
The second and third child instances must be newly purchased — they cannot be converted from existing instances.
For the full list of restrictions, see Limits of Global Distributed Cache.
Convert an existing instance
This path converts an existing Tair memory-optimized instance into the first child instance of a new distributed instance.
Conversion causes a transient connection interruption of a few seconds. To minimize impact:
Perform the conversion during off-peak hours.
Make sure your application has a reconnection mechanism.
Log in to the ApsaraDB for Redis console.
In the left navigation pane, click Global Distributed Cache.
In the upper-left corner, click Create Instance > Create From Existing Instance.
In the panel that appears, select the instance to convert into the first child instance. We recommend that you set Execution Method to Convert Within Maintenance Window.
Click OK.
After conversion, the instance is no longer listed in the standard instance list. Manage it by clicking Global Distributed Cache in the left navigation pane.
Purchase a new child instance
This path creates a new Tair memory-optimized instance and enrolls it as the first child instance of a distributed instance. Use this path if you don't have an existing instance to convert.
Log in to the ApsaraDB for Redis console.
In the left navigation pane, click Global Distributed Cache.
In the upper-left corner, click Create Instance > Create Distributed Instance.
On the purchase page, configure the following settings:
Setting Options Notes Service Tair (Enterprise Edition) Fixed — no other options Billing method Subscription / Pay-as-you-go Subscription: pay upfront, cost-effective for long-term use. Pay-as-you-go: billed hourly, suitable for short-term use. Deployment mode cloud-native (recommended) / classic Cloud-native is currently in public preview. If the option is unavailable, fill out the request form to get access. Storage medium Memory Fixed — no other options Region and kvstore_zone_group Select based on your setup Select the same region and zone as your Elastic Compute Service (ECS) instance for lower latency, or select a region close to your on-premises device. VPC and vSwitch Select based on your setup Use the same Virtual Private Cloud (VPC) as your ECS instance to allow internal network communication. Different vSwitches within the same VPC can still communicate. Set password Now Set the instance password immediately. Subscription duration Choose a duration Only applies to the Subscription billing method. Click Buy Now.
On the Confirm Order page, read the Terms of Service and complete the payment.
After payment, wait 1–5 minutes. The child instance is ready when its status changes to Running.
What's next
FAQ
Why do I get the error "The gdc custins trans not support"?
The instance you're trying to convert is a classic deployment mode instance with a direct connection address enabled. Release the direct connection address, then retry the conversion. Cloud-native instances are not affected by this limitation.
API reference
| API operation | Description |
|---|---|
| CreateTairInstance | Creates a memory-optimized instance to use as the first child instance or add to an existing distributed instance as the second or third child instance. |
| CreateGlobalDistributeCache | Converts a specified instance into the first child instance of a distributed instance. |
| DescribeGlobalDistributeCache | Queries detailed information about a distributed instance. |
| RemoveSubInstance | Removes a child instance from a distributed instance and converts it back to a standard instance. Data is preserved. |