Detailed Explanation Of Cross-Region Backup And Disaster Recovery Implementation For Alibaba Cloud’s Singapore Servers

For those deployed in Alibaba Cloud Singapore servers For operations in (ap-southeast-1), when designing cross-regional backup and disaster recovery solutions, one often faces a trade-off among three criteria: “most stable (best)”, “most suitable (optimal)”, and “lowest cost (cheapest)”. To achieve the highest availability and shortest RTO, it is recommended to use active hot standby/dual-active (cross-regional synchronous replication) in combination with DTS Real-time database synchronization, cross-regional load balancing, and CEN Private network interconnection ； If cost-effectiveness is a priority, you can choose to transfer the ECS image/EBS snapshot through HBR Or archive OSS in China or other regions as cold backup, and restore as needed ； To achieve balance (the best outcome), adopt a “warm standby” architecture: Key databases used DTS For continuous incremental replication, files are replicated across regions using OSS, along with lifecycle management. ECS uses regular snapshots for this purpose HBR Management, combining to achieve controllable costs and short to medium RTOs.

Common Alibaba Cloud components used for implementing cross-regional disaster recovery include: ECS (compute nodes), cloud disks, and ECS snapshots, HBR (Hybrid Backup Recovery) is used for backup management OSS Used for object storage and cross-region replication DTS Used for real-time database migration/replication, CEN Or VPC Peering is used for network connectivity, Cloud DNS is used for traffic switching, SLB (Load Balancing) is used for traffic distribution, and CloudMonitor and Log Service are used for monitoring and audit of drills.

Cold backup: Backup data is periodically uploaded to OSS Or a snapshot repository in the target area; resources need to be started during recovery, resulting in a long RTO but the lowest cost ； Warm-up: Reserve basic resources in the target area or use on-demand elastic resources; continuous replication of critical data (DTS), with moderate RTO and moderate cost ； Hot standby/dual-active: Simultaneous external service provision at both locations or hot standby failover, using real-time replication and traffic scheduling; it results in the shortest RTO but the highest cost. It is recommended to classify by business type: use warm or hot backups for core transactions/databases, and cold backups for static files and logs.

For relational databases (such as RDS/MySQL/PolarDB), it is recommended to use built-in backups in combination with DTS Perform asynchronous real-time or near-real-time replication: First, perform a full synchronization, then enable incremental/log replication (binary log or CDC), and maintain a read-only instance or standby instance in the target area. Regularly verify incremental consistency, set appropriate binlog retention periods and network bandwidth limits to ensure that RPO meets business requirements. For non-relational or self-built databases, logical backups combined with periodic archiving using HBR or OSS can be used, while DTS can replace or supplement real-time synchronization.

At the ECS host level, ECS snapshots can be used in conjunction with HBR Management. Recommendation strategy: Frequent incremental snapshots are taken of the critical hosts (e.g., every hour or every 4 hours), and these are replicated by HBR to the target region/backup repository. The operating system/application configurations are also included in the image management ； For files and media content, OSS is preferred. Cross-Region Replication should be enabled, along with lifecycle rules, to archive cold data to low-frequency/archival storage in order to save costs.

Cross-regional disaster recovery requires the ability to quickly redirect traffic in the event of a failure. Common Practices: Use CEN Or use VPC Peering to establish inter-regional private links, reducing cross-domain public network traffic and link instability ； Combine SLB with cloud DNS (Cloud DNS/DNS from Alibaba Cloud) to achieve DNS-level switching, or use CDN/Global Accelerator for cross-regional traffic distribution. When switching, session persistence and reduced TTL for certificates and domain name resolution must be considered to speed up the switch.

During design, first classify the business by level (e.g., P0-P3): For P0 (transactional) systems, with RPO ≤ 5s and RTO ≤ 5 minutes, real-time replication + hot backup should be used ； P1 (core service): RPO of a few minutes, RTO of several dozen minutes; uses warm backup + DTS + snapshots ； P2 (General Services): RPO of several hours, RTO of several hours; uses HBR periodic snapshots + OSS archiving ； P3 (logging/audit) can accept an RPO of several days, using cold backup. After defining the SLA, allocate backup frequency, data retention, and drill frequency by tier.

Cross-regional transmission and backup must consider data security and compliance. It is recommended to enable encryption for backup data both during transmission and at rest: Use Alibaba Cloud KMS to manage keys, enabling server-side encryption (SSE) and transport encryption (HTTPS) for HBR and OSS. For sensitive data, consider also data masking, minimizing cross-border data transfers, and complying with local regulations (such as the compliance requirements of Singapore and the country where the business is located). At the same time, set up access control, MFA, and audit logs to ensure that backup operations have an audit trail.

The effectiveness of disaster recovery plans is ensured through regular drills and monitoring. Create automated drill scripts (backup and recovery, DNS failover, database master-slave failover), conduct failure drills during off-peak hours, and record RTO/RPO differences. Monitor backup success rate, DTS latency, snapshot failures, and target region resource status through CloudMonitor, Log Service, and alert policies. Develop a detailed Runbook (fault diagnosis, switching steps, failback procedures, contact list) and update it regularly.

Suggestions for controlling cross-regional DR costs: 1) Tier data storage (hot/cold/archived) and use OSS lifecycle policies ； 2) Use incremental snapshots and deduplication compression to reduce storage usage ； 3) Utilize on-demand reserved instances/elastic instances and auto-scaling to avoid long-term costs ； 4) Do not activate redundant computing resources during non-drill periods; only retain data copies ； 5) Evaluate cross-regional bandwidth and outbound costs, prioritizing dedicated lines/CEN to reduce long-term costs and improve stability.

1. Business classification and RTO/RPO definition ； 2. Select the target area (can be domestic or another overseas area), and confirm the availability zone and pricing ； 3. Set up network connectivity (CEN or VPC Peering) and test network latency and bandwidth ； 4. Configure full + incremental replication of database DTS or built-in RDS backup to the target area ； 5. Configure ECS snapshot policy and connect to HBR for replication to the target zone ； 6. Configure OSS cross-region replication and lifecycle rules ； 7. Configure DNS/SLB failover policy and certificate synchronization ； 8. Write a runbook and conduct drills to optimize processes ； 9. Online monitoring, alerts, and auditing.

For Alibaba Cloud Singapore servers Designing cross-regional backup and disaster recovery should take into account business importance, cost tolerance, and compliance requirements. To pursue “the best,” one chooses active-active/hot standby and real-time replication; to pursue “the cheapest,” one tends to use cold backup plus OSS archiving. “The optimal” approach is usually a warm backup strategy combined with the use of DTS/HBR/OSS as needed. Regardless of which approach is chosen, the key is to define RTO/RPO, automate testing and monitoring, to ensure rapid recovery in accordance with SLAs in the event of a real failure.