This topic describes the vertical and horizontal elastic scaling modes for Hologres compute groups. This topic also explains how to use these scaling capabilities. You can choose a scaling method based on your requirements.
Elastic scaling modes
Hologres lets you divide an instance into multiple compute groups. For each compute group, you can choose one of the following scaling directions based on your requirements:
Vertical scaling (scale-up/down): Starting with Hologres V2.0, you can vertically increase or decrease the computing resources of a compute group. This mode is suitable for both medium-to-large task scenarios and high-concurrency, small-task scenarios.
Horizontal scaling (scale-out/in): Starting with Hologres V3.2, you can horizontally increase or decrease the number of clusters in a compute group. This mode is better suited for high-concurrency, small-task scenarios and provides more flexible throughput and isolation capabilities.
Scaling methods
You can use the scaling capabilities in the following ways:
Manual scaling: You can manually perform vertical scaling and adjust the number of clusters. For more information, see Compute group management.
Time-based scaling (vertical): You can set scheduled tasks for vertical scaling. For more information, see Time-based scaling (Beta).
Auto scaling (horizontal): You can set a scaling limit to automatically scale out based on the load. For more information, see Multi-cluster and auto scaling (Beta).
Glossary
The following terms are used in the time-based scaling feature:
Compute group reserved resources: The amount of resources reserved for a compute group. You can define this amount when you create the compute group and modify it later. These resources are allocated from the instance's reserved resources.
Compute group elastic resources: The amount of resources that are dynamically scaled out in addition to the reserved resources using the time-based scaling mechanism.
Total computing resources of a compute group: The total amount of resources used by the compute group. This is the sum of the compute group reserved resources and compute group elastic resources.
The following terms are used in the multi-cluster and auto scaling feature:
Reserved cluster count: The number of clusters reserved for a compute group. You can define this number when you create the compute group and modify it later. The computing resources for reserved clusters are allocated from the instance's reserved resources.
Specifications of a single cluster: The specifications for a reserved cluster. You can define these specifications when you create the compute group and modify them later. These specifications determine the amount of resources that each reserved cluster is allocated from the instance's reserved resources.
Compute group reserved resources: Specifications of a single cluster × Reserved cluster count.
Compute group elastic resources:
The amount of resources scaled out in addition to the reserved resources using auto scaling.
Specifications of a single cluster × (Current cluster count - Reserved cluster count).
Total computing resources of a compute group: The total amount of resources used by the compute group. This is the sum of the compute group reserved resources and compute group elastic resources.
Instance-level computing resources are defined as follows:
Instance reserved resources: The total amount of computing resources reserved for an instance. These resources are available for both subscription and pay-as-you-go billing methods. These resources are divided into allocated resources and unallocated resources:
Instance allocated resources: The portion of instance reserved resources that is allocated to all compute groups as their reserved resources.
Instance unallocated resources: The portion of instance reserved resources that is not allocated to any compute group.
Instance elastic resources: The total amount of resources scaled out for all compute groups in an instance, in addition to their reserved resources. This scaling is performed using time-based scaling or auto scaling.
Total computing resources of an instance: The total amount of resources used by the instance. This is the sum of the instance reserved resources and instance elastic resources.