0

There are multiple distribution in the company. Building the same project results in different build times (see table 1).

| win10laptop32gbram | win10laptopOld | win7laptop | win7server | linuxvm | mac |
|--------------------|----------------|------------|------------|---------|-----| 
| 3                  | 3              | 2          | 1.5        | 1       | 1.9 | 

Table 1: Ratio of build speed between different Operating System and hardware.

Do other people from other organizations experience this as well?

2
  • 4
    Are your build done on the same hardware? If not there's no point in comparing build time for OS comparison at all asthey are biased by hardware
    – Tensibai
    Mar 31, 2018 at 18:03
  • 1
    Slower at doing what? It's almost certainly going to depend on what exactly you're doing, and "in general" is way too broad to be able to answer effectively. Apr 5, 2018 at 16:50

1 Answer 1

7

TL;DR depends on the use case with up to 12 times between best/worst

Comprehensive testing has been done over at Phoronix (28.03.18) and the results are very mixed. The tests looked at:

  • Clear Linux 21510
  • Debian 9.4
  • OpenSUSE 42.3 Leap
  • Ubuntu 16.04 LTS
  • Windows 10 Pro Build 16299
  • Debian 9 On WSL
  • OpenSUSE 42.3 On WSL
  • Ubuntu 16.04 LTS On WSL

The results are all over the show. In a few tests, the WSL (Windows Subsystem for Linux) is faster then native Windows (GraphicsMagick) and the Intel optimized distribution, Clear Linux, sometimes has the worst results (Netperf UDP request response). While in others Windows is twice as fast as all Linux distributions (CacheBench). All in all, deciding on performance is not conclusive. Choose the OS that provides the tooling you require.

Nonetheless, the article itself does provide a winner given that only first place finishes are taken into account:

... Clear Linux was the fastest of the operating systems tested with coming in first 40% of the time, Debian 9.4 in second with first place finishes 22.5% of the time, and Windows 10 Pro itself was the fastest 12.5% of the time.

When only taking Java into account:

In the Java tests, to no real surprise, the performance tended to be about the same across all tested operating systems.

And finally, in a compile test (Golang), which is the test that is most related to the use case in the question:

The build performance on Windows 10 and WSL were slower than the bare metal Linux performance.

2
  • Could you the conclusions of the various paragraphs to prevent that the outcomes of the research will vanish if the link is deprecated. I will upvote the answer right now and if you could add the six conclusions, I will accept the answer.
    – 030
    Apr 1, 2018 at 13:11
  • This is interesting, but with WSL, I assume this is using GCC on windows. I would be interested in also seeing a comparison between Visual Studio's engine and Borland as well with GCC and GCC running on windows and Linux. Apr 2, 2018 at 16:55

Not the answer you're looking for? Browse other questions tagged or ask your own question.