Today I did a small test. I tried using our Linux Kernel compile script on the Intel Xeon E5-2630L V3 in two different modes. First was bare metal, second was using a Docker image. Both were Ubuntu Xenial based and the script times compile not any of the pre-requisites. Nothing else was running on the machine.terminal
It turns out that doing the linux kernel compile benchmark was 12.5% slower.
Each one takes >10 minutes and we use the fifth run to let the cores sufficiently heat up and get some heat soak. Even looking at the record of the first four compiles it seems to be the same difference.
I remember when we were profiling applications for Linux-Bench most were very close (+/- 2%), but there were a few that were significantly slower in Docker. That was the major reason for not using Docker with Linux-Bench.
Has anyone else found applications that are running significantly slower in Docker than on bare metal?
It turns out that doing the linux kernel compile benchmark was 12.5% slower.
Each one takes >10 minutes and we use the fifth run to let the cores sufficiently heat up and get some heat soak. Even looking at the record of the first four compiles it seems to be the same difference.
I remember when we were profiling applications for Linux-Bench most were very close (+/- 2%), but there were a few that were significantly slower in Docker. That was the major reason for not using Docker with Linux-Bench.
Has anyone else found applications that are running significantly slower in Docker than on bare metal?
