Amazon, Clear, Debian, Gentoo, Red Hat, SUSE & Ubuntu Performance On The EC2 Cloud

[phoronix]

Phoronix: Amazon, Clear, Debian, Gentoo, Red Hat, SUSE & Ubuntu Performance On The EC2 Cloud

It's been a few months since last running a Linux distribution / operating system comparison on Amazon's EC2 public cloud, but given the ever-advancing state of Linux, here are some fresh benchmarks when testing the Amazon Linux AMI, Clear Linux, Debian 9.2, Gentoo, Red Hat Enterprise Linux 7.4, SUSE Linux Enterprise 12 SP3, and Ubuntu 16.04 LTS.

http://www.phoronix.com/vr.php?view=25583

[andrei_me]

One advantage of virtual environment, no firmware time spent on boot

After various post about Clear Linux perf advantages, I'm reading their documentation to see how it works and will try on a spare server

[starshipeleven]

One advantage of virtual environment, no firmware time spent on boot

For those wondering why this person brings up this as an advantage, server boot times are awfully long, like multiple minutes long, just because the firmware takes a long time to initialize the damn thing.

[andrei_me]

tildearrow you appear as "Honorary Editor" here in the forums, congratulations, well deserved title (y)

[reavertm]

It's incorrect to call this distro a "Gentoo" in this benchmark. Gentoo is a meta-distribution. What is being benchmarked here is just one of instances. Gentoo instances differ from each other with CFLAGS/CXXFLAGS/LDFLAGS, package.use, etc. It's a bit like calling Ubuntu "Debian". More appropriate name would be just Pygoscelis Papua Linux.

[dkasak]

It's incorrect to call this distro a "Gentoo" in this benchmark. Gentoo is a meta-distribution. What is being benchmarked here is just one of instances. Gentoo instances differ from each other with CFLAGS/CXXFLAGS/LDFLAGS, package.use, etc. It's a bit like calling Ubuntu "Debian". More appropriate name would be just Pygoscelis Papua Linux.

Agreed. While I understand Mike would probably not be too keen to do a custom build just for this kind of benchmark, these kind of comparisons are really only benchmarking selected packages compiled with different versions of gcc and different CFLAGS. I wonder how many people running Gentoo are using gcc-5.4? Maybe we should build Mike an AMI for future Gentoo tests?

[andrei_me]

Michael as Amazon bills per hour, how long it took to complete all these tests?

[Michael]

Michael as Amazon bills per hour, how long it took to complete all these tests?

~12 hours

[reavertm]

Agreed. While I understand Mike would probably not be too keen to do a custom build just for this kind of benchmark, these kind of comparisons are really only benchmarking selected packages compiled with different versions of gcc and different CFLAGS. I wonder how many people running Gentoo are using gcc-5.4? Maybe we should build Mike an AMI for future Gentoo tests?

"Most" of people running Gentoo were using gcc-5.4[1] I assume. We stabilized gcc-6.4 only recently[2].
That being said, while having some Gentoo box benchmarked in Phoronix is nice as it gives the distro some exposure, everyone needs to be aware it's not indicative of general "Gentoo" performance.
Gentoo is what you make it to be. It can be made one of the fastest or the slowest.
Benchmark in article is surely indicative though about Pygoscelis Papua Linux as this is some pre-packaged and pre-configured instance and as such tests are reliably reproducible.
Building distro with default USE flags (and make.conf in general) could be indicative of default "Gentoo", but frankly I don't see much benefit for such benchmarks myself.

1. https://packages.gentoo.org/packages/sys-devel/gcc
2. https://gitweb.gentoo.org/repo/gento...37e988e2c47a16

[PeeJay]

One of us running Gentoo should write an article benchmarking common variations of cflags and compiler versions. The "fast" settings would likely end up being similar to clear Linux I guess.