Hologres:Select an instance type for Hologres

Instance type

Resource type

Core requirements

Scenarios

Virtual warehouse instance (Recommended)

Dedicated computing and storage resources

• High-performance, high-concurrency, and large-scale data processing.

• A single instance can implement a read/write splitting architecture.

• Large-scale data services with data volumes in the terabyte (TB) or petabyte (PB) range and frequent complex queries.

• High-concurrency online services that require millisecond-level responses (over 1,000 QPS).

• Mixed payload isolation, such as isolating online analytical processing (OLAP) queries from real-time writes, or isolating resources for different lines-of-business.

• Scenarios with scalability requirements, such as significant traffic fluctuations with periodic peaks that require rapid scaling.

Serverless instance

Shared computing resources and dedicated storage resources

• High-performance, large-scale, and low-frequency data manipulation and queries.

• Development or staging environments.

• Scenarios with low-frequency, high-performance queries.

• Scenarios where business payloads fluctuate significantly, peaks are hard to predict, and you cannot reserve dedicated computing resources in advance.

General-purpose instance

Dedicated computing and storage resources

• Cost control and lightweight analytics.

• Can be upgraded to a virtual warehouse instance.

• Small data volumes (less than 1 TB).

• Lightweight analytics, such as low-frequency ad hoc queries and small- to medium-scale business intelligence (BI) report generation.

• Cost-sensitive businesses with limited budgets that require a cost-effective solution. Suitable for trial phases of non-core services.

• Low-concurrency services, such as internal data service platforms and scheduled task processing.

Read-only replica instance (To be discontinued)

Dedicated computing and storage resources

• A multi-instance read/write splitting architecture.

• A read-only replica instance is used by attaching it to a general-purpose instance.

• To be discontinued. We recommend that you use virtual warehouse instances instead.

• The general-purpose primary instance is used for data writes and data manipulation. The read-only replica instance is used for data analytics to ensure read/write splitting.

Shared cluster (Lakehouse Acceleration Edition) (To be discontinued)

Shared computing resources and no storage resources

• Low-cost acceleration for data lakehouse queries and BI integration.

• No need to read from or write to Hologres internal tables.

• Scenarios where data is stored in OSS or MaxCompute and you want to use Hologres to accelerate queries.

• Scenarios with large burst queries where reserved dedicated instance resources are insufficient.

• Scenarios where you connect to BI tools to accelerate queries on data lakehouse data. You can use Hologres shared clusters to connect to BI tools such as Tableau, Quick BI, and FineReport.