Linux KPTI Tests Using Linux 4.14 vs. 4.9 vs. 4.4

Written by Michael Larabel in Software on 4 January 2018. Page 2 of 3. 30 Comments

With the simple SQLite benchmark, having KPTI disabled saves about one second off the ~30 second test run, but there was no real difference between the kernel versions tested and fall within the margin of error.

With x264 video encoding, no real difference in performance caused by KPTI but Linux 4.14 is running slightlyu faster than the older kernels.

With the ebizzy benchmark that aims to resemble web server workloads, no real changes to point out.

In the timed Apache compilation process, using Linux 4.14 takes about one second longer to build than Linux 4.9 or 4.4, but no change caused by KPTI and still within the margin of error.

While in the timed kernel compilation process the results ended up being about the same with just about an extra second tacked on for the KPTI builds. In our tests on the other systems, the slowdown with KPTI when building the Linux kernel tended to be more pronounced.

With PostgreSQL, Linux 4.14 overall is the fastest in this databsse benchmark and this is one of the I/O workloads where Kernel Page Table Isolation does cause a performance increase. But the relative performance for the three tested kernels with KPTI on/off were about the same with no significantly different performance out of the older kernels lacking PCID optimizations.

Related Articles