Clear Linux Ending Out 2018 With Even More Performance Optimizations

Written by Michael Larabel in Operating Systems on 29 December 2018 at 03:06 PM EST. Page 1 of 2. 15 Comments.

With the Windows Server 2019 vs. Linux benchmarks this week on a dual socket Intel Xeon Scalable server and testing six different Linux distributions and three Windows Server configurations, Intel's open-source Clear Linux was the winner in nearly half of the dozens of benchmarks carried out across these Linux and Windows operating system tests. But the results did yield some areas they could improve upon for better performance and as a result have already landed some more performance optimizations.

As a result of the recent Windows Server vs. Linux benchmarks, at least one takeaway was the Rscript performance not meeting their expectations. To which they have now pushed out an update that resolves an issue in their OpenBLAS linear algebra library use by R as well as Numpy and other packages. With this fixing/tuning, the performance should be back to their expectations. Curious, I ran some benchmarks to see the difference.

I did load up Clear Linux 26930 on the same dual Xeon Gold 6138 server where the recent Windows Server vs. Linux benchmarks happened.

Windows Server 2019 vs. Linux OS Benchmarks On Dual Intel Xeon Server

Indeed, the R benchmark was much faster... The original Clear Linux performance already was faster than Windows Server and Debian/Ubuntu, but now it squeezed out better performance. Though surprisingly Fedora Server 29 still managed to have slightly better performance.

The other tests not building against OpenBLAS obviously weren't affected with this targeted change besides the usual rolling changes on Clear:

Windows Server 2019 vs. Linux OS Benchmarks On Dual Intel Xeon Server
Windows Server 2019 vs. Linux OS Benchmarks On Dual Intel Xeon Server
Windows Server 2019 vs. Linux OS Benchmarks On Dual Intel Xeon Server

I next ran some tests on a Xeon Silver box...


Related Articles