Tenant Cloud Storage Cloud Object Storage Best Practices

Disaster Recovery and High Availability Architecture Based on Replication Within the Cloud

Last Updated At: 2025-10-21 09:10:00

Overview

COS provides customers with 99.95% availability and 99.999999999% reliability. However, due to many uncontrollable factors such as natural disasters and fiber optic failures, neither the availability nor the reliability of cloud data can be 100%. Meanwhile, for certain industries, such as the financial industry, high availability and high reliability need to be ensured for their business due to the special nature of their business.

In order to achieve business continuity and stability and meet the needs of enterprises for high availability and high reliability, COS provides a data disaster recovery and high-availability solution based on the inter-cloud replication feature. We recommend that when enterprises use the cloud, they perform disaster recovery backup of cloud data according to business needs to ensure continuous and stable business operations.

This document mainly introduces two aspects. First, it introduces a disaster recovery solution for primary/replica switching of cloud business based on inter-cloud replication. On the other hand, it further introduces a high-availability solution based on inter-cloud replication, which achieves high business availability through inter-cloud replication, origin-pull, SCF, CDN, and other products and features.

DR and Backup Solutions Based on Inter-Cloud Replication

Disaster recovery needs to meet three elements: Redundance, Remote, and Replication.

  • Redundance: That is, data redundancy, which requires that data be backed up to another available system at the same time.
  • Remote: Refers to the storage of backup data in another region that is far away. Because disasters often have geographical continuity, only a sufficiently long distance can guarantee the availability of redundant data.
  • Replication: Refers to zero loss of backup data.
    COS's inter-cloud replication feature can achieve inter-cloud synchronization of incremental data. The data uploaded by users can be copied to a storage bucket in another region within a few seconds to tens of minutes, depending on the file size and the distance between regions. Based on inter-cloud replication, data can be backed up in different locations for redundancy, thus achieving business disaster recovery. For an introduction to inter-cloud replication, see Inter-Cloud Replication Overview. To enable inter-cloud replication, you need to enable the versioning feature first. For an introduction to versioning, see Versioning Overview.

The following figure shows the architecture diagram of DR and backup based on inter-cloud replication.

In this architecture, the customer's storage bucket A and storage bucket B serve as the primary/replica of each other. In this architecture, the customer's storage bucket A and storage bucket B serve as the primary/replica of each other. To ensure business continuity and stability, the enterprise configured inter-cloud replication rules for bucket A and bucket B, respectively. When the inter-cloud replication rule is in effect, the incremental data in bucket A will be automatically copied to bucket B, and the incremental data in bucket B will also be automatically copied to bucket A.

Note:

After the incremental data in bucket A is copied to bucket B, although it is the incremental data in bucket B, it will not be copied to bucket A again.

Under normal circumstances, the enterprise's primary read/write request links all points to bucket A, and all incremental data will be automatically and incrementally synchronously copied to bucket B as backup data. The client can add a network quality detection module to the upload or download program. When the primary bucket A is detected to be down, the read and write request link can be quickly switched to the replica bucket B.

High-Availability Solution Based on Inter-Cloud Replication

The above document introduces a disaster recovery backup solution based on inter-cloud replication. This solution can use existing products and features on the cloud to implement data backup and disaster recovery switching. However, the actual business operation status may be complex and diverse, and the above-mentioned disaster recovery backup solutions may not be able to guarantee the high availability of the business. Therefore, this section proposes a high-availability solution based on inter-cloud replication, which achieves high business availability through inter-cloud replication, origin-pull, SCF, CDN, and other products and features.

The following figure shows the architecture diagram of business HA based on inter-cloud replication.

This architecture is mainly divided into the following levels:

  • HA layer: Integrates network detection and business scheduling and switches links based on metrics, such as link connectivity. Users can implement it based on SCF (refer to the introduction in the previous section) or implement it on the client based on business requirements.
  • Storage layer: Generally, it consists of buckets of COS in different regions.

The way this architecture ensures high availability of business is described as follows:

  1. Under normal circumstances, the enterprise's primary write request link points to bucket A, and all incremental data will be automatically and synchronously copied to bucket B as backup data.
  2. When the link to primary bucket A is not accessible (for example, the dial quality decreases or upload failure is detected), the client can switch the write request link to primary bucket B. At this time, all incremental data will also be automatically and synchronously copied to bucket A.
  3. Customers can also choose to back up redundant data on their own origin server or other cloud vendors and configure an origin-pull policy for bucket B. Assume that in an extreme case, the links of primary buckets A and B are both unavailable at the same time. Then, if uploading data to bucket B fails, bucket B can pull data from the origin server.

Note:

  • The cost of a full redundant backup of data is high. Customers can also choose to only redundantly back up hot data (such as files uploaded within a few hours) to reduce data storage costs.
  • If you choose the origin server as part of your high-availability architecture, be careful to assess the origin server bandwidth and the impact of its limitations when designing the architecture.

Reference Document

The following documents may help you implement a disaster recovery and highly-availability architecture: