Vector Retrieval Using Compressed Vectors to Cut Shared Storage Delay
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vector retrieval solutions, particularly in shared storage scenarios, suffer from high delays due to synchronous disk I/O operations and network I/O costs, leading to increased latency and reduced user experience.
Innovation Solution
Store compressed vectors in a high-performance first-level storage resource and original vectors in a lower-performance second-level resource, using a pre-trained classification model to identify candidate vectors from the first-level resource, followed by similarity calculations on the second-level resource to reduce data access and volume.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Volume of stationary object
If original vectors are stored in a second-level storage resource with lower data access performance, then storage capacity is improved, but vector retrieval delay increases
Solution Approach 1:
The patent segments the storage system into two levels: a first-level storage resource for compressed vectors and a second-level storage resource for original vectors. This segmentation allows the system to leverage the fast access of the first level for preliminary filtering while using the high-capacity second level for final retrieval, thus resolving the contradiction between storage capacity and retrieval speed.
Solution Approach 2:
The patent introduces compressed vectors as an intermediary representation between the query and the original vectors. These compressed vectors serve as a bridge that enables fast preliminary similarity calculations in the first-level storage, reducing the number of expensive disk I/O operations needed to access the second-level storage.
2Productivity
If compressed vectors are used for preliminary filtering, then the number of data accesses to second-level resource is reduced, but system complexity increases
Solution Approach 1:
The patent changes the parameter representation of vectors by creating compressed versions that retain essential similarity information while occupying less space. This parameter transformation enables efficient preliminary filtering in the first-level storage without requiring complex additional hardware or infrastructure.
Data Source
AI summary
This application provides vector retrieval methods, and apparatuses, and storage media. In an example, in response to a query vector, at least one compressed vector having a highest similarity with the query vector is obtained from a first-level storage resource. Classification decision-making is performed on each compressed vector based on a classification model to obtain at least one target compressed vector whose corresponding original vector is a candidate vector most similar to the query vector. At least one target original vector corresponding to the at least one target compressed vector is obtained from a second-level storage resource. At least one similarity between the query vector and the at least one target original vector is calculated. The at least one target original vector is sorted based on the similarity, based on which, a vector retrieval result corresponding to the query vector is determined.