Shared vs. dedicated cluster comparison
Shared cluster instances are cost-effective because they use shared resources. Use them for scenarios that require fewer storage and compute resources.
Dedicated cluster instances have exclusive access to all resources. They support more algorithm optimization features and industry templates. Use them for scenarios with high resource consumption or that require high-quality search results.
Price comparison
The pricing for Shared and Dedicated clusters is different. For specific pricing, see the product purchase page.
System limit comparison
Feature list | Item | Shared cluster | Dedicated cluster |
Search-related | Maximum length of each clause (excluding filter) | 1 KB before encoding | Configurable as needed. (Contact technical support.) |
Maximum length of a filter clause | The first 4 KB before encoding | Configurable as needed. (Contact technical support.) | |
Maximum number of results for search retrieval | 5,000 | Configurable as needed. (Contact technical support.) | |
Number of documents for rough sort | 1,000,000 | Configurable as needed. (Contact technical support.) | |
Number of documents for fine sort | 200 by default | Configurable as needed. (Contact technical support.) | |
Accurate statistics for the aggregate clause | Accuracy is guaranteed for fewer than 100,000 records. | Accuracy is guaranteed for more than 100,000 records. (Note: Set enable_accurate_statistics:true in the kvpairs clause to enable accurate statistics.) | |
Sort policy | Sort policy configuration | Rough sort + Fine sort | Coarse and Fine Sorting |
Rough sort type | Weight expression | Weight and formula expressions | |
Maximum number of sort policies | 30 | 50 | |
Advanced custom sorting plugin - Cava | Not supported | Supported | |
Application schema | Multilayer association of primary and secondary tables | Up to two layers | Up to three layers |
Instance creation/upgrade or downgrade | Creation method | Direct activation | Applications are deployed separately. If you exceed the automatic deployment quota, contact technical support. |
Scaling | Flexible scaling | Applications are deployed separately. If you exceed the automatic deployment quota, contact technical support. | |
Industry template | E-commerce Industry Algorithm Edition | Not supported | Free upgrade to the E-commerce Industry Algorithm Edition. |
Education Industry Algorithm Edition | Not supported | Free upgrade to the Education Industry Algorithm Edition. | |
Game Industry Algorithm Edition | Not supported | Free upgrade to the Game Industry Algorithm Edition. | |
Content Community Industry Algorithm Edition | Not supported | Free upgrade to the Content Community Industry Algorithm Edition. |
Performance comparison
Feature list | Item | Shared cluster | Dedicated cluster |
Data synchronization | Total API request capacity per second | 2 MB/s | 15 MB/s by default. Configurable as needed. (Contact technical support.) |
Maximum rate for incremental synchronization from RDS/PolarDB | 2 MB/s | 15 MB/s by default. Configurable as needed. (Contact technical support.) | |
Maximum total TPS for primary table updates and primary table updates triggered by secondary tables | 1,500 | Configurable as needed. (Contact technical support.) | |
Retrieval | Query latency | Millisecond-level | Millisecond-level |
The automatic deployment quotas for Dedicated cluster applications are: Dedicated General-purpose (150 GB, 2,000 Lightweight Compute Units (LCUs)), Dedicated Compute-optimized (120 GB, 3,000 LCUs), and Dedicated Storage-optimized (750 GB, 1,500 LCUs). If you exceed the quota, contact technical support for deployment assistance.