Linux Patches Boost Per-Core I/O Performance by 60%

Newly submitted patches for the Linux kernel are showing a 60% improvement in per-core I/O performance, a breakthrough that could significantly enhance how the operating system handles storage workloads. The proposed changes target the block layer, the part of the kernel responsible for managing data transfers between memory and storage devices.

The performance gains come from optimizing how the kernel schedules and dispatches I/O requests on a per-CPU-core basis. Currently, locking overhead and contention between cores can slow down throughput, especially on systems with many processors. The patch set, authored by kernel developer Ming Lei, introduces a lockless per-core I/O completion path that reduces this bottleneck. Early benchmarks reveal that the new code nearly doubles the input/output operations per second (IOPS) on multi-core machines, with a particularly pronounced effect on fast NVMe storage.

This is not just a theoretical improvement. For data centers running Linux servers, a 60% boost in per-core I/O could translate into faster database queries, smoother virtual machine operations, and more efficient handling of large-scale file systems. The patches are currently under review for inclusion in a future kernel release, likely targeting the Linux 6.13 cycle.

While the patches are still experimental, the results align with a broader push within the Linux community to reduce kernel overhead and make better use of modern multi-core hardware. As storage speeds continue to outpace CPU clock rates, eliminating inefficiencies in the I/O path becomes critical. If these patches are merged, they could represent one of the most impactful performance optimizations for Linux storage in recent years.