Multi-zone (Basic Edition) deployment

更新时间:
复制 MD 格式

The Multi-zone Basic Edition deploys Lindorm across multiple data centers to prevent service unavailability from data center-level failures. Compare its architecture and cost with the Multi-zone High-availability Edition to choose the right deployment option.

Background information

A single-zone Lindorm instance deploys all data and master nodes in one availability zone (AZ). If a data center failure makes that AZ unavailable, the entire Lindorm service becomes unavailable.

A Multi-zone (Basic Edition) instance distributes master and data nodes across multiple data centers. If a data center-level failure occurs in one zone, data integrity and partial service availability are maintained. The level of availability depends on the computing and storage resources in the remaining zones.

The two editions differ as follows:

  • Multi-zone (Basic Edition): Storage replicas are stored across data centers, but there is only one set of computing resources. This requires some recovery time in the event of a failure. The Recovery Time Objective (RTO) is within 1 hour, but this edition is more cost-effective.

  • Multi-zone (High-availability Edition): Each data center contains independent storage, services, and two sets of computing resources. If a failure occurs, you can switch directly between the primary and secondary clusters. The RTO is less than 3 minutes.

Multi-zone (Basic Edition) architecture

image

The Multi-zone (Basic Edition) architecture ensures high availability in the following ways:

  • Cross-data center deployment of master nodes: All master nodes are deployed across three data centers to ensure metadata high availability. This prevents split-brain issues if a network partition occurs.

  • Flexible distribution of compute nodes: For cloud disk deployments, compute nodes are distributed across two zones. For local disk deployments, they are distributed across three zones. Reserve at least 30% redundant resources for CPU computing and storage space to prevent a single-zone failure from affecting system availability.

  • Upgraded OSS deployment mode: OSS uses a zone-redundant storage deployment mode, ensuring stable cold storage services and high data availability.

High availability comparison

The following table compares availability, performance, and cost across Lindorm single-zone, Multi-zone (Basic Edition), and Multi-zone (High-availability Edition) instances.

Item

Single-zone

Multi-zone (Basic Edition)

Multi-zone (High-availability Edition)

Number of cross-AZs

1

3

3

Supported engines

All

All

For LindormTable only

RPO and RTO during a data center failure

No data center-level disaster recovery

  • Recovery Point Objective (RPO)=0

  • RTO < 1 hour (with sufficient resources)

  • RPO=0

  • RTO < 3 minutes

Cost

1

  • 1× to 1.25×

  • No cost change for standard storage types (such as standard cloud storage and performance-optimized cloud storage).

  • 25% increase for storage-optimized cloud storage.

1.5× to 2×

Performance

1

  • Significant increase in single-row write latency (0.2 ms -> 1.5 ms).

  • Single-row query time may increase due to cross-data center access.

  • Significant increase in single-row write latency (0.2 ms -> 1.5 ms).

  • Query latency has almost no attenuation (due to nearest access).

The Multi-zone Basic Edition provides data center-level disaster recovery at minimal additional cost.

Important
  • Because the Basic Edition has only one set of computing resources, computing and storage capacity drops significantly during a data center failure. Maintain sufficient resource overhead to ensure capacity during disaster recovery.

    • For a cloud disk deployment, computing and storage capacity decreases by 50% during a single data center failure.

    • For a local disk deployment, computing and storage capacity decreases by 30% during a single data center failure.

  • The Multi-zone (Basic Edition) writes data to multiple data centers by default, which may increase the write response time (RT). To mitigate this, use batch writes to significantly reduce the perceived increase in write RT.

  • If you upgrade from a single-zone instance to a Multi-zone (Basic Edition) instance, response time (RT) may change. Monitor your services closely during the upgrade.

Limits

  • The Multi-zone (Basic Edition) is currently available only in the following regions:

    • China: China (Ulanqab) General Industry Cloud, China (Qingdao), China (Hangzhou), China (Beijing), China (Heyuan), China (Shanghai), China (Zhangjiakou), China (Shenzhen), China (Chengdu), China (Hong Kong), China (Guangzhou), China (Ulanqab), and China (Hohhot).

    • Asia Pacific: South Korea (Seoul), Japan (Tokyo), Singapore, Malaysia (Kuala Lumpur), Indonesia (Jakarta), and Thailand (Bangkok).

    • Other: US (Atlanta), UAE (Dubai), Germany (Frankfurt), US (Silicon Valley), UK (London), and US (Virginia).

  • The cloud disk data nodes of a Multi-zone (Basic Edition) instance are evenly distributed across two data centers. Therefore, when you create an instance or scale out nodes, the number of nodes must be a multiple of 2.

  • The local disk data nodes of a Multi-zone (Basic Edition) instance are evenly distributed across three data centers. Therefore, when you create an instance or scale out nodes, the number of nodes must be a multiple of 3.

Purchase a Multi-zone (Basic Edition) instance

You can purchase a Multi-zone (Basic Edition) instance in the console.

FAQ

Why does the console show only one zone for a multi-zone instance?

For multi-zone instances, the Lindorm console currently shows only the primary zone. The secondary zone and the client node zone are not displayed.