PolarDB:Enable serverless for IMCI read-only nodes

Prerequisites

Your PolarDB for MySQL cluster is Enterprise Edition, runs kernel version 8.0, and has at least one IMCI read-only node.

Implementation options

• Set a resource scaling range on a fixed-specification IMCI read-only node: This method adds serverless capabilities to a node that has a fixed specification. The fixed specification provides a baseline processing capacity, and resources can temporarily scale up to handle pressure when the workload exceeds this baseline.

• Create an IMCI read-only node with no fixed specification: This method creates a node with no fixed specification. Its resources scale on demand based on the workload, which is ideal for cost-sensitive use cases.

Benefits

Cost reduction for peak and off-peak cycles

A serverless architecture excels at handling fluctuating workloads. It automatically scales down resources during off-peak periods and scales up during peak periods, effectively reducing costs. For example, consider a workload that requires a 32-core, 64 GB specification for 12 peak hours and only a 16-core, 32 GB specification for 12 off-peak hours. A serverless configuration based on the 16-core, 32 GB specification can reduce compute node costs by 25% compared with a fixed 32-core, 64 GB specification.

Optimizations for the serverless architecture

IMCI provides the following optimizations for elastic resources in a serverless architecture to further improve performance and the user experience:

• Optimized resource utilization: IMCI read-only nodes feature a comprehensive resource management mechanism that supports online configuration of key system parameters, such as cache, thread pools, and execution memory. In a serverless architecture, these parameters are automatically adjusted when resources scale up or down to ensure maximum resource utilization.

• SQL queuing optimization: IMCI read-only nodes are primarily used for analytical workloads. To run complex SQL queries quickly, the system typically allocates more resources to each complex query. When the number of complex queries increases, system resources can become constrained. This can block simple queries submitted later, causing slow queries due to queuing. In a serverless architecture, IMCI read-only nodes automatically add resources based on SQL queuing conditions to reduce slow queries.

  Note

  The following kernel versions support this optimization:

  • If the major version of PolarDB for MySQL is 8.0.1, the revision must be 8.0.1.1.39 or later.

  • If the major version of PolarDB for MySQL is 8.0.2, the revision must be 8.0.2.2.20 or later.

• Faster IMCI loading on startup: After a cluster restart, an IMCI read-only node must rebuild its internal index metadata, which typically takes several minutes. During this time, IMCI is unavailable. In a serverless architecture, the IMCI read-only node uses elastic resources to accelerate this rebuilding process, making it available sooner.

  Note

  The following kernel versions support this optimization:

  • If the major version of PolarDB for MySQL is 8.0.1, the revision must be 8.0.1.1.36 or later.

  • If the major version of PolarDB for MySQL is 8.0.2, the revision must be 8.0.2.2.20 or later.

• Accelerated log replay: IMCI is built by replaying physical logs. However, because the system allocates most of its resources to support queries, resources for building the IMCI can be insufficient. This can increase the IMCI log replay latency. This latency can route queries to other nodes or force them to wait for the log replay to complete, resulting in slow queries. In a serverless architecture, an IMCI read-only node automatically adds replay-related resources to speed up log replay.

  Note

  This optimization is supported if the major kernel version of PolarDB for MySQL is 8.0.1 and the revision is 8.0.1.1.45 or later.

• Accelerated DDL operations: DDL operations on IMCI can be time-consuming. In a serverless architecture, IMCI nodes adaptively increase the thread resources for DDL processing to accelerate the operation.

  Note

  The following kernel versions support this optimization:

  • If the major version of PolarDB for MySQL is 8.0.1, the revision must be 8.0.1.1.44 or later.

  • If the major version of PolarDB for MySQL is 8.0.2, the revision must be 8.0.2.2.25 or later.

Procedure

Set scaling range for fixed-spec nodes

1. Log in to the PolarDB console。

2. In the upper-left corner, select the region where the cluster is deployed.

3. Find the target cluster and click its ID.

4. On the Basic Information page, in the Database Nodes section, click Enable Serverless.

5. In the Enable Serverless dialog box, set Maximum Resources for Single Node and Minimum Resources for Single Node. For more information about Maximum Resources for Single Node and Minimum Resources for Single Node, see Enable the serverless feature for a fixed-specification cluster.

6. Click OK.

   After you apply the settings, the IMCI node can scale dynamically in addition to its fixed specification.

   In the Database Nodes section, the IMCI read-only node's specification is displayed as a fixed specification plus a PCU scaling range, such as 4 cores 8 GB + (0–16) PCU. The node status is Running.

Create a node without fixed specifications

1. Log in to the PolarDB console。

2. In the upper-left corner, select the region where the cluster is deployed.

3. Find the target cluster and click its ID.

4. On the Basic Information page, in the Database Nodes section, click Enable Serverless.

5. In the Enable Serverless dialog box, specify the number for Read-only IMCI Nodes and click OK. This example creates one IMCI read-only node.

   After you apply the settings, PolarDB automatically creates an IMCI read-only node that dynamically scales from 1 to 16 PCU.

   After the node is created, you can view the new IMCI read-only node in the Database Nodes section on the Connection and Node Management page. Its status is Running.