KIOXIA Updates AiSAQ Software for AI Vector Search Optimization
KIOXIA today announced an update to its KIOXIA AiSAQ (All-in-Storage ANNS with Product Quantization) software, an open-source tool designed to improve AI vector database searches within retrieval-augmented generation (RAG) systems by optimizing the use of solid-state drives (SSDs). The new version introduces flexible controls, allowing system architects to adjust the balance between search performance and the number of vectors on fixed-capacity SSD storage without hardware modifications.
First released in January 2025, KIOXIA AiSAQ software employs an approximate nearest neighbor search (ANNS) algorithm optimized for SSDs, eliminating the need to store index data in DRAM. This approach enables vector searches directly on SSDs and reduces host memory requirements, allowing vector databases to scale with fewer restrictions related to DRAM capacity.
The balance between increasing search performance (queries per second) and maximizing the number of vectors on a fixed SSD capacity varies by workload. The latest KIOXIA AiSAQ software update provides configurable options to achieve this balance across different RAG system workloads. KIOXIA states that this makes the technology suitable for RAG applications and other vector-intensive applications, such as offline semantic searches.
SSDs offer an alternative to DRAM for managing the high throughput and low latency required by RAG systems, driven by the demand for scalable AI services. KIOXIA AiSAQ software aims to meet these demands by enabling large-scale generative AI without being constrained by limited memory resources. KIOXIA’s decision to release AiSAQ software as open-source reinforces its commitment to promoting SSD-centric architectures for scalable AI within the AI community.
Axel Störmann, Vice President and Chief Technology Officer for Memory and SSD products, KIOXIA Europe GmbH, stated, “We are always looking to help developers and system architects to fine-tune performance and capacity in innovative new ways. Now, with the introduction of the latest version of KIOXIA AiSAQ software, end users can depend on the power of SSDs to build scalable RAG systems flexibly and efficiently. By open-sourcing our technology, we are reaffirming our unwavering commitment to the AI community and giving them solutions that are both powerful and accessible.”
The KIOXIA AiSAQ open-source software is available for download.
