Verified, then logged.
Sun 22 Nov 2026 · 00:38 KST / 15:38 UTC한국어 | ENSearch /

08 / Industry · Analysis

KAIST filesystem cuts delays in storage data processing

Software changes require no extra hardware, but storage processing gains must be read separately from overall AI system efficiency.

An open storage server reveals rows of SSDs and neatly routed cables beside a laptop displaying a text-free storage-management interface in a data center. AI-generated image

KAIST researchers developed a filesystem with up to threefold faster data processing, AI Times reported on 11 Nov 2026. KAIST is a South Korean science and technology university. The software, called GAZE, manages data on storage devices.

AI Times linked the gains to software changes requiring no extra hardware. The work targets bottlenecks when reading and writing data for AI. Its reported scope concerns storage, rather than overall AI performance.

AI Times identified AI systems and data centres as potential beneficiaries. It described better storage operating efficiency as an expected benefit. That distinction separates the processing gains from their potential use.

Reducing delays during data cleanup

AI Times attributed the work to Professor Youjip Won's team at KAIST. The team is in the School of Electrical Engineering. It focused on delays during data cleanup in next-generation flash storage.

According to AI Times, GAZE reduces unnecessary waits and processing burdens. The reported improvement comes from changing storage management software. How storage is managed is therefore central to the development.

The limits of the reported processing gain

AI Times described the threefold improvement as a maximum. That wording does not establish the same gain across all workloads. Storage processing performance is also distinct from overall AI performance.

AI Times linked the expected efficiency benefit to fewer storage bottlenecks. Its account presents data centre operating efficiency as a potential outcome. The remaining distinction is between faster processing and gains in wider use.