How the New Linux Swap Table Boosts Performance by Up to 20%

Tencent engineer Kairui Song recently submitted a new patch to the Linux kernel mailing list, proposing a new architecture called Swap Table that integrates the swap cache, swap mapping, and swap allocator into a unified backend infrastructure.

The Phase I patch contains nine changes and has already shown clear benefits. In tests on both small ARM devices and large x86_64 servers, performance improvements were observed:

Overall acceleration of roughly 5%–20%.

In virtual‑machine scalability scenarios, some tests exceeded a 20% improvement.

Kernel build time reduced by several percentage points.

Throughput of in‑memory databases such as Redis and Valkey increased by 6%–7%.

This optimization not only delivers higher throughput and lower latency but also reduces memory consumption, laying a solid foundation for future extensions and feature enhancements.

The first round of Swap Table patches released in May already demonstrated 20%–30% performance gains; the current patch further strengthens this potential.

Although some developers argue that modern systems often minimize or disable swap, the community broadly acknowledges the value of this improvement for high‑load and virtualized environments.

As the patches continue to evolve, the Swap Table optimization is expected to become the default mechanism in future Linux kernels, providing more stable and efficient memory management.