All Products
Search
Document Center

OpenSearch:Specification calculator

Last Updated:Jul 23, 2024

OpenSearch Vector Search Edition provides the specification calculator feature to help you configure instances that better meet your business requirements. To use the specification calculator, click Create Instance on the Instances page of the OpenSearch Vector Search Edition console. The specification calculator appears on the right of the buy page.

Click Specification Calculator. In the OpenSearch Vector Search Edition Specification Calculator dialog box, configure the following parameters.

Basic configurations

  1. The region in which the instance resides: the region in which you want to create the OpenSearch instance.

  2. Whether Disaster Recovery is Required: specifies whether disaster recovery is required. Valid values: Yes and None.

Vector data configurations

  1. Vector Data Entries: the number of documents in the vector data that you want to write to the instance.

  2. Vector Dimension: the dimensions of the vector that you want to write to the instance.

  3. Vector Algorithm: You can select a vector algorithm based on your business requirements. The following three vector algorithms are supported:

  1. HNSW: The HNSW algorithm is based on proximity graphs and provides a high retrieval rate and good performance on low-dimensional and high-dimensional vector datasets. The memory and storage usage of the HNSW algorithm are comparable to those of the Linear algorithm. It is applicable to most vector retrieval scenarios.

  2. QC: The QC algorithm is based on quantitative clustering and ensures high accuracy of retrieval results. It uses only one-fourth of the memory and storage space required for the HNSW or Linear algorithm and provides relatively good performance on low-dimensional vector datasets. It is applicable to the scenarios in which a large amount of data needs to be retrieved and a high retrieval rate is not required.

  3. Linear: the full-text retrieval. The retrieval results are completely correct, but the algorithm requires a large amount of resources and provides poor performance. It is applicable to accurate retrieval on small datasets that contain no more than 10,000 data records.

Query configurations

  1. Average QPS: the queries per second (QPS) of the instance.

  2. Expected Average Time of Response to Query: the average time that is used to retrieve results. You can select a unit from the drop-down list. Valid values: s and ms.

After the preceding configurations are complete, click Perform Operation.

The specification calculator provides a recommendation on the specifications of Query Result Searcher (QRS) workers and searcher workers and the number of replicas.

You can purchase an instance based on the recommendation of the specification calculator.