Mastering Hologres Compute Groups: Isolation, Elasticity, and Serverless Strategies

Hologres Compute Group Instance Overview

The real‑time data warehouse faces four main resource challenges: load isolation, resource waste, large‑task OOM issues, and operations difficulty.

Hologres addresses these with elastic compute group instances (V2.0) that separate storage and compute, allowing multiple isolated compute groups to share a single data store.

Load Isolation (V2.0)

Each compute group has physically isolated resources, shares the same data and metadata, and can be created, scaled, or restarted independently. A single endpoint routes SQL to the appropriate group via a gateway.

Time‑Based Elasticity (V2.2)

Elastic plans let you reserve baseline CU resources and add elastic CU during peak periods, improving utilization and reducing cost by up to 30%.

Serverless Computing (V2.1)

Serverless Computing uses a separate resource pool to handle large write or query tasks, avoiding OOM and providing stable execution for big jobs.

Query Queue (V3.0)

Query Queue offers load throttling and large‑query isolation, helping solve operational challenges.

Compute Group Practical Exercises

1. Compute Group Management

Demonstrates creating, scaling, and rebalancing compute groups, showing resource allocation (e.g., a 64 CU instance split into two 32 CU groups).

2. Compute Group Authorization

Shows how users are granted table permissions and how compute groups are explicitly authorized for resources and table groups, including leader/follower settings and replica configuration.

3. Connecting to Compute Groups

Explains default connection using the standard endpoint and explicit connection by appending @warehouse_name to the database name in JDBC/PSQL strings.

4. Load Isolation Demo

Illustrates separate workloads (1 billion row write vs. TPC‑H Q1 query) running on different compute groups, confirming isolation via monitoring metrics.

5. Monitoring Metrics

Describes viewing CPU, memory, QPS, latency, and I/O metrics per compute group.

6. Converting Instances to Compute Group Type

Outlines constraints (minimum 32 CU, version ≥2.0.4) and steps for converting general or primary‑secondary instances to compute‑group instances, including downtime considerations.

Time‑Based Elasticity Practical Exercises

1. Concept and Billing

Explains reserved vs. elastic resources and cost savings when using elastic plans during peak hours.

2. Elasticity Configuration

Shows how to set elastic plans for specific time windows (e.g., 2 am‑6 am with 32 CU elastic) and how changes take effect immediately if the current time falls within the plan.

3. Monitoring and Alerts

Details monitoring elastic core usage, execution logs, and cloud monitoring events for elastic scaling actions.