Before You Start: High Availability and Reliability Requirements

The DWS service logic consists of the service layer, GaussDB(DWS) cluster layer, OS layer, and VM layer, as shown in Table 1.

The performance of your GaussDB(DWS) service system is impacted by these logical layers. In the event of service changes or hardware faults, you may need to make temporary adjustments or perform emergency recovery.

To ensure efficient use of GaussDB(DWS) clusters and achieve a quick recovery time objective (RTO), adhere to the rules provided in Table 2 when setting up the system.

**Table 1** GaussDB(DWS) service logic layers
No.	Logic Layer	Description	Service Characteristic
1	Service layer	Applications on the service side	Service applications: gsql, jdbc, odbc, python, datastudio, and navicat.
2	GaussDB(DWS) layer	GaussDB(DWS) clusters	Service application logic is distributed to each CN in the GaussDB(DWS) cluster via SQL statements. Each CN then queries, optimizes, and forwards the data to DNs for processing. After processing, the DNs aggregate the data and return it to the CNs, which then send it back to the application.
3	OS layer	OS installed for GaussDB(DWS) clusters	This layer provides the running foundation, OS, file system, and network services for GaussDB(DWS).
4	VM layer	VM where GaussDB(DWS) clusters are located	This layer provides EVS disks, networks, and resources like CPU and memory.

**Table 2** Cluster high availability and reliability risks and suggestions
Cluster Configuration	Recommended Measure	Risk If Not Configured
Configuring load balancing	Use ELB load balancing to access services. For details, see Binding and Unbinding ELBs for a GaussDB(DWS) Cluster.	Services are unavailable when the CN is faulty.
Backing up configuration data	Use the backup function to create redundant copies of both upstream and downstream data.	Data cannot be restored after being deleted by mistake, which affects your service usage and causes data security faults.
Configuring cluster DR	Configure cluster or service DR. For details, see GaussDB(DWS) Cluster DR Management.	In disaster scenarios at the cluster level, it is not possible to guarantee service continuity.