How to choose an instance type for AnalyticDB for PostgreSQL - AnalyticDB

This topic describes how to choose an instance type for AnalyticDB for PostgreSQL.

Instance resource types

AnalyticDB for PostgreSQL offers two recommended instance resource types: Storage-elastic Mode and Serverless.

Storage-elastic Mode

This mode provides a full range of features and uses an integrated architecture for compute and storage. It supports scaling compute nodes vertically or horizontally and allows for flexible storage adjustments.

When you purchase an instance, you must specify the Instance Edition, Compute Node Specifications, Compute Nodes, storage node type, and storage size per node.
Serverless

Built on a proprietary architecture that decouples compute and storage, Serverless supports scaling in seconds and on-demand storage. It is ideal for workloads with highly variable resource demand.

When you purchase an instance, you must specify the Instance Edition, Compute Node Specifications, and Compute Nodes.

The following tables provide more details.

Table 1. Storage-elastic Mode

Instance family	Node specifications	Recommended disk types	Recommended use cases
High Availability	2 cores, 16 GB	ESSD cloud disk PL0	Proof-of-concept (POC) testing. Personal learning and evaluating product features.
	4 cores, 32 GB	ESSD cloud disk PL0 ESSD cloud disk PL1	A popular choice for balanced compute and storage needs, selected by 60% of users.
	8 cores, 64 GB	ESSD cloud disk PL1	Suitable for compute-intensive scenarios, such as complex data analytics and high-concurrency workloads.
	16 cores, 128 GB	ESSD cloud disk PL2	Recommended for building large-scale, enterprise-level data platforms that handle high-concurrency workloads.
High-performance (Basic Edition)	2 cores, 8 GB	ESSD cloud disk PL0	POC testing. Personal learning and evaluating product features.
	4 cores, 16 GB	ESSD cloud disk PL0 ESSD cloud disk PL1	Suitable for offline analytics workloads with balanced compute and storage needs. Important The High-performance (Basic Edition) family does not provide high availability. Choose this family with caution.
	8 cores, 32 GB	ESSD cloud disk PL1
	16 cores, 64 GB	ESSD cloud disk PL2

Table 2. Serverless

Instance family	Scheduling mode	Node specifications / ACU threshold	Storage disk type	Recommended use cases
High Availability	Manual Scheduling	4 cores, 16 GB	Shared Storage	This new instance type supports scaling in seconds, on-demand storage, and data sharing. Suitable for the following scenarios: Workloads with high variability. New services with unpredictable resource requirements. Workloads that require clear resource isolation.
		8 cores, 32 GB
	Automatic Scheduling (In invitation-only preview. You can Submit a ticket to apply for access.)	8 to 32 ACU

Instance configuration use cases

Use case 1: Internet and manufacturing companies

You are an Internet or manufacturing customer who currently uses a self-managed database and a Greenplum data warehouse and wants to migrate your deployment to the cloud.

Recommendation: Deploy AnalyticDB for PostgreSQL in Storage-elastic Mode.

Benefits: AnalyticDB for PostgreSQL is fully compatible with Greenplum and PostgreSQL syntax and their open source ecosystems. This allows a seamless migration and lets you adjust resources on demand.

Use case 2: SaaS providers

You are a Software-as-a-Service (SaaS) provider building a data platform that ingests data from various sources like ApsaraDB RDS, Flink, and OSS. You need to perform ETL processes, support mixed analytical workloads, and ensure high availability. Your platform must also provide data for reporting and other enterprise data services.

Recommendation: Use AnalyticDB for PostgreSQL in Storage-elastic Mode. Select the High Availability instance family with compute node specifications of at least 4 cores and 32 GB, and a minimum of four compute nodes.

Benefits: Storage-elastic Mode integrates with the Alibaba Cloud ecosystem, allowing you to import data from other Alibaba Cloud services. It provides enterprise-grade features like workload management through UDFs or resource queues. You can perform entire ETL workflows within the data warehouse. It delivers up to three times the compute performance of traditional data warehouse systems, supports vertical and horizontal scaling, and offers excellent platform extensibility to adapt to changing analytics needs.

Use case 3: Traditional enterprises undergoing digital transformation

You are a traditional enterprise undergoing digital transformation and need to replace a legacy on-premises data warehouse, such as Teradata, Oracle, DB2, or Greenplum.

Recommendation: Use AnalyticDB for PostgreSQL in Storage-elastic Mode. Choose the High Availability or High-performance (Basic Edition) family based on your business needs. We recommend compute node specifications of at least 4 cores and 32 GB, with a minimum of four compute nodes.

Benefits: AnalyticDB for PostgreSQL is a proven replacement for Teradata and Oracle, with its capabilities and solutions validated by hundreds of customers in finance, telecommunications, and government sectors.

Use case 4: Autonomous driving companies

You are in the autonomous driving industry and need to analyze geospatial and time-series data collected from vehicles. You require strong support for the JSON format, spatiotemporal data analytics capabilities, business intelligence dashboards, and feature engineering.

Recommendation: Use AnalyticDB for PostgreSQL in Storage-elastic Mode. Select the High-performance (Basic Edition) instance family with compute node specifications of at least 4 cores and 32 GB.

Benefits: Storage-elastic Mode includes the PostGIS and Ganos spatiotemporal analytics engines and accelerates queries with its MPP architecture. It supports analysis of semi-structured data like JSON and provides data lake analytics for maximum flexibility.

Use case 5: Online gaming companies

You are an online gaming company building a data platform to analyze user behavior data. The platform cleans business logs, performs data correlation analysis, and provides real-time support for operational tools. You have mixed workloads during business hours and need resource isolation.

Recommendation: Use AnalyticDB for PostgreSQL Serverless. We recommend compute node specifications of at least 4 cores and 16 GB, with a minimum of four compute nodes.

Benefits: Serverless allows you to flexibly adjust resources based on different time periods. It offers a mature solution for log data using Log Service (SLS) and OSS, enabling efficient in-warehouse data cleaning. Serverless provides comprehensive analytics capabilities and powerful single-node compute performance.

Use case 6: New retail businesses

You are a new retail business building a Customer Data Platform (CDP). Your platform needs to ingest data from multiple sources and provide a mature solution for downstream audience segmentation.

Benefits: Storage-elastic Mode supports various data formats, such as JSON, CSV, AVRO, and PARQUET, enabling rapid data aggregation and tag generation. It integrates with Alibaba Cloud's proprietary products like Quick Audience, allowing you to quickly build an all-in-one cloud platform.

Use case 7: Large internet enterprises

You are a large internet enterprise where different business lines have their own data platforms, but the company also has a unified central data platform. You want to quickly deploy independent resources for different business workloads without creating data silos in the future.

Recommendation: Use AnalyticDB for PostgreSQL Serverless. Deploy multiple instances with compute node specifications of at least 4 cores and 16 GB, and a minimum of two compute nodes per instance.

Benefits: The rapid resource deployment and elasticity of Serverless eliminate the need for extensive upfront resource planning, allowing you to dynamically adapt to business workloads. Data can be shared across multiple instances, which prevents the formation of data silos as business platforms evolve. Independent instances ensure complete resource isolation, and usage for each business line can be tracked directly through billing.

Use case 8: Building a data development platform

You need to build a data development platform and want to minimize the impact of development activities on online production services while increasing development efficiency.

Benefits: Use AnalyticDB for PostgreSQL Serverless for your production database. When you need to perform data development, you can create a test instance and use data sharing to access a real-time, synchronized copy of production data. This approach prevents development work from affecting the production environment while allowing developers to work with the most up-to-date data.