Linux 6.8 Network Optimizations Can Boost TCP Performance For Many Concurrent Connections By ~40%

[sophisticles]

Can we please finally get a method to ignore this moron?

Translation, you don't have hard numbers.

As I have said, Linux use is a religion, and like all religions, the adherents get angry when you ask them to back up their claims with actual facts.

[sophisticles]

That's a well-known fact... even Microsoft Azure has more Linux usage than all of the other operating systems combined.

Pretend that it's not that well known and show me some actual data.

[sophisticles]

What?! Linux was always the most performant and FreeBSD was also good. Mentioning Windows and Mac Server is some kind of joke? They're nowhere near Linux TCP performance (and not only TPC). Some server benchmarks:

https://interworks.com/blog/tkau/2017/09/13/tableau-server-linux-performance-sneak-peek/

http://dev.squaresolutions.net/pcaps/2019093/test03-h_upload_linux/test03-h_upload_linux-throughput.png

http://dev.squaresolutions.net/pcaps...throughput.png

it's not even funny, but make me wonder where did you came from?

Did you bother to read the linked article?

Note: As with any performance scenario testing, these results are exclusive to this test, and the objective was to hold everything constant except the version of Tableau Server. While I expect the directional result “Linux is faster” to hold true, these exact times will not be reproducible in real life or in your own testing.

The entire test is invalid and the author even acknowledges that.

[nadirx]

Pretend that it's not that well known and show me some actual data.

https://azure.microsoft.com/en-us/products/virtual-machines/linux

More than 60 percent of customer cores in Azure run Linux workloads. Choose from popular Linux distributions including Red Hat, SUSE, Ubuntu, CentOS, Debian, and CoreOS. Use preconfigured solutions from Oracle and other open-source virtual machine (VM)-compatible providers and find Azure-optimized Linux VM images from publishers of your choice. Get the cloud's only co-located technical support from Azure, Red Hat, and SUSE with just one ticket. Choose the underlying processors_AMD, Arm64, or Intel_that best meet your requirements.

[nadirx]

He admits he conducted a half-assed test and in the end both OSes where able to fully saturate the NIC.

And in your mind this shows that Linux is better than BSD?

You should go back and read Matteo's article without just extrapolating a couple of pieces here and there:
* He concludes that FreeBSD's VirtIO driver is terrible, so he goes with a hardware implementation to overcome this. All the numbers after this are not "half-assed"
* Linux is much better at bridging interfaces and scales with multiple cores
* Linux routing scales much better with multiple cores

You keep asking for proof and numbers from others, but carefully avoid providing any numbers yourself, which, according to your own statements, disqualifies you from participating in the discussion.

[coder]

From that article:

"Both OS performs well, being able to forward more than 1 million of pps per core, which lets you achieve the 10 gbit line rate with 1500 byte frames."

Yes, this was never in doubt, IMO. However:

1. The industry has moved well beyond 10 Gbps, in case you haven't noticed.
2. Not all network traffic is using max MTU, such as in the case of certain server apps.
3. CPU overhead still matters, as we've seen from the original article to which this thread is attached. The higher performance you can achieve over loopback, the lower your CPU overhead should be when transmitting at a lower, line-limited speed.

From that article:

"So I did the test again on the loopback interface on both guests. It’s not a very professional test, but at least I’m not facing any VirtIO deficience."

As he explains, the reason he tested this way was to avoid putting FreeBSD at an artificial disadvantage. It wasn't done to make Linux look better. Were that the objective, he could've just posted his VirtIO-based results and called it a day.

And in your mind this shows that Linux is better than BSD?

I'd have rather he used bare hardware than VMs, but the data does show Linux with substantial advantages in many areas. You should be aware that networking code is modular and it's not as if merely using a different interface changes out the entire network stack for something different. So, I accept the results as a reasonable approximation of the two OS' capabilities.

If you have better data, let's see it.

In Michael's article, the various Linux and BSD distros trade punches across the benchmarks but it doesn't really matter.

"trade punches" makes it sound like they're evenly matched. They're clearly not.

However, your statement was concerning network performance and that's unfortunately not what Michael's testing focused on.

Despite what the people of this forum, at least the vocal ones, seem to think, raw performance is not as important as reliability and uptime in a server.

That's changing the subject. It's not a bad subject, but the abrupt shift is kinda sus...

Anyway, I care a lot more about data than anecdotes. If you have statistically significant reliability data to share, feel free to.

[coder]

Translation, you don't have hard numbers.

As I have said, Linux use is a religion, and like all religions, the adherents get angry when you ask them to back up their claims with actual facts.

I could say the same thing about you and your religious anti-Linux/FOSS bigotry. Your claims lack supporting data. You attack data that's provided, without offering anything better, and then try to change the subject when it's clear you've lost.

[Volta]

I've exported the latest performance data between Mac Server, FreeBSD and Linux:

You've exported your brain.

[Volta]

The entire test is invalid and the author even acknowledges that.

We've got a moron here. Yes, I read the article and unlike you I understand what they said. In short words: there are many tunables, but in their tests Linux destroyed windows by a large margin. Btw. did you miss another one I gave in the same comment?

[Volta]

Translation, you don't have hard numbers.

As I have said, Linux use is a religion, and like all religions, the adherents get angry when you ask them to back up their claims with actual facts.

You have claims you can't prove. We've showed you many evidences, so shut up.