# Researchers Boost SSD Speeds by Resolving OS Bottlenecks New GAZE filesystem technology triples flash storage performance by optimizing garbage collection processes. - **Source:** [Digital Trends](https://www.digitaltrends.com) - **Published:** 2026-11-11 - **Topic:** [technology](https://brifing.si/c/technology) - **Original article:** - **On Brifing:** https://brifing.si/s/28261f61-2504-462e-a6cc-f1c5c5abebe6 ## Flash Storage Performance Limits Modern flash storage relies on garbage collection to reuse space by moving valid data and wiping old sectors. Researchers at KAIST identified that current operating systems struggle to manage this process efficiently on Zoned Namespace storage. This architectural mismatch creates significant latency, preventing high-speed drives from reaching their full potential during intensive operations. ## The ZNS Cleanup Bottleneck Zoned Namespace storage divides drives into gigabyte-sized zones, shifting cleanup responsibilities to the filesystem. Testing revealed that cleaning a single 1 GB zone took 1.7 seconds, which is 283 times slower than smaller sections. Surprisingly, 70% of this delay stems from CPU overhead rather than the physical storage hardware limitations itself. ## Introducing the GAZE Filesystem The KAIST team developed GAZE to resolve these OS-level inefficiencies without requiring specialized hardware. By splitting cleanup into four overlapping steps, the system reads new data batches while simultaneously writing previous ones. It also pre-allocates memory for cleanup tasks, effectively removing the need for drives to wait for piece-by-piece memory requests. ## Performance Gains and Efficiency GAZE delivers up to three times the performance of standard setups and previous solutions like ZNS+ or P-GC. The technology reduces total cleanup time to approximately one-third of its original duration. Because it functions as a software-based filesystem modification, it integrates seamlessly into existing storage architectures without needing any physical drive changes. ## Impact on AI Infrastructure This breakthrough is particularly critical for AI inference tasks, where processors frequently lose efficiency while waiting for data. By eliminating the operating system bottleneck, GAZE allows data centers to utilize the full speed of their existing hardware investments. This ensures that high-performance chips remain fed with data, reducing idle time significantly. ## Broader Context of Storage Flash drives cannot overwrite data directly, necessitating the housekeeping process known as garbage collection. As storage densities increase, the complexity of managing these zones grows, making filesystem-level optimizations essential. The GAZE approach demonstrates that software-defined improvements can often outperform hardware-only upgrades when addressing systemic bottlenecks within modern computing environments and server architectures. ## Research Presentation Details Led by Professor Youjip Won, the research team presented their findings at the ACM SOSP 2026 conference held in Prague. The study highlights a fundamental shift in how operating systems interact with modern flash storage. By focusing on CPU-bound tasks, the team successfully addressed the primary cause of storage latency in ZNS configurations. ## Future Implications for Data The successful implementation of GAZE suggests a path forward for scaling storage performance in cloud and enterprise environments. By removing the OS as a limiting factor, developers can focus on maximizing throughput for data-intensive applications. Future adoption could redefine how data centers manage massive datasets, ensuring faster access speeds for all users. ## Sources covering this story - https://www.digitaltrends.com/computing/kaists-gaze-filesystem-could-make-next-gen-ssds-up-to-3x-faster/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._