A data engineer is designing a disaster recovery strategy for a database system that serves a geographically dispersed user base. The business requires that users in different regions access a local copy of the database to minimize latency and improve performance. Data consistency across all regions is important but can tolerate slight delays (seconds to minutes). Which replication strategy is most suitable for this scenario?
- AMulti-Master Replication
- BSynchronous Replication
- CSnapshot Replication
- DMaster-Slave Replication
Show answer & explanationAnswer & explanation
Correct answer: A. Multi-Master Replication
Multi-Master (or Active-Active) replication allows multiple database instances in different geographical locations to accept write operations. This enables users to connect to a local database for both reads and writes, minimizing latency. Data consistency is maintained across all masters, typically with asynchronous propagation, which naturally accommodates 'slight delays' while providing continuous availability and localized performance.
Why the other options are wrong
- B. Synchronous replication would introduce unacceptable latency for geographically dispersed users trying to write to a 'local' copy that must wait for remote commits.
- C. Snapshot replication is for periodic full copies and would not support continuous availability with local read/write access and near real-time consistency.
- D. Master-Slave replication would require all writes to go to a single master, negating the benefit of local write access for geographically dispersed users and centralizing write latency.
Multi-Master Replication
Multi-Master replication (also known as Active-Active replication) is a database architecture where multiple database servers (masters) can accept read and write operations. Changes made on one master are replicated to all other masters, allowing for load balancing, high availability, and reduced latency for geographically dispersed users.
- Multiple nodes can accept read and write operations.
- Improves performance by allowing local writes for dispersed users.
- Requires robust conflict resolution mechanisms and tolerates slight replication delays.
Memory trick: Multi-Master: Everyone writes, everyone reads, globally distributed needs.