Linux kernel v7.0 introduces a substantial speedup in reclaiming system memory that is used for caching large files, targeting scenarios where servers keep large datasets in RAM to avoid frequent storage access. According to notes on the kernel mailing list, a new set of patches queued for the Linux 7.0 merge window showed reclaim speed improvements of up to 75% in testing. This focuses on improving the efficiency of freeing cached file data when memory pressure rises, aiming to reduce stalls and latency during cleanup.
In one benchmark, developers allocated 10 GB of file-backed data in memory and then reclaimed 8 GB of it. On a 32-core Arm64 server, the reclaim process completed about 75% faster compared to the older Linux implementation, while on an x86 machine, the improvement was reported at over 50%. These results highlight that both Arm64 and x86 architectures benefit meaningfully from the new reclaim logic, with especially strong gains on many core Arm64 servers commonly used in cloud environments.
The optimization is particularly relevant for systems running large databases or other memory intensive services, where the kernel may keep tens or even hundreds of gigabytes of frequently accessed file data in RAM for faster access. When memory pressure builds and some of that cached data must be freed, the cleanup process now finishes significantly faster, reducing the impact on running workloads. While typical consumer systems are unlikely to notice a difference, hyperscalers, high performance computing simulations, Artificial Intelligence runs, and other heavy data processing workloads can see a significant performance boost. The improvement was authored by Baolin Wang of Alibaba, who focused on optimizing how the kernel handles large blocks of cached file memory.
