Virtual Cluster Mirroring – One of Seven Core Capabilities of GBase 8a V9

The virtual cluster mirroring feature is a key product capability of GBase 8a MPP v9.5 virtual clusters. It enables real-time active-active operation between two logical sub-clusters, achieving T+0 high availability for data. Through mirroring technology, data can be replicated between different virtual clusters, providing enhanced disaster recovery capabilities and meeting requirements such as flexible system disaster tolerance and read/write workload separation.

Virtual cluster mirroring can establish a mirroring relationship between two logical sub-clusters at the database or table level. Once the mirroring relationship is established, the data in the two mirrored tables is synchronized in real time, cluster-level table locks are applied to both tables simultaneously, and the two mirrored tables act as mutual backups.

The following example illustrates how to create and use virtual cluster mirroring. In a virtual cluster, there are two logical sub-clusters: VC1 and VC2. VC1 contains the database VC1.db1, and VC2 contains the database VC2.db2. Provided that the two databases, db1 and db2, have the same data distribution strategy, a mirroring relationship can be created between VC1.db1 and VC2.db2. After the relationship is established, db1 and db2 will have identical tables and data.

The operations on databases with a mirroring relationship are briefly described as follows:

1. DDL operations issued by the cluster layer are simultaneously sent to db1 and db2 for execution;

2. For DML operations issued by the cluster layer, db1 performs the computation, and the results are sent to the secondary shards of db1 as well as the primary and secondary shards of db2 using the data forwarding and replication mechanism of GBase 8a MPP Cluster;

3. DQL operations issued by the cluster layer are executed only on one of the databases, db1 or db2.

Now that you have a basic understanding of virtual cluster mirroring, let's summarize its key features:

● High Availability: With a mirroring relationship, data tables gain additional replicas, significantly improving database high availability;

● Active-Active: Virtual cluster mirroring enables real-time active-active operation between two logical sub-clusters, meeting the demands of flexible disaster recovery and read/write workload separation;

● Intra-City Disaster Recovery: Virtual cluster mirroring supports deploying clusters across different locations within the same city, enabling intra-city disaster recovery.