Originally posted by Michael_S
View Post
Announcement
Collapse
No announcement yet.
Facebook Is Hiring To Make Linux Networking Better Than FreeBSD
Collapse
X
-
Originally posted by xeekei View PostWhile you guys throw fanboy poop at each other, all I have to say is: thank you Facebook for contributing to open source software. It'll be interesting to see what comes out of this effort.
More seriously, I agree with your point that it's good Facebook is doing this.
Originally posted by xeekeiAll you needed to do was look at Facebook's past projects before posting. Google "HipHop PHP".
I'd prefer GPLv3, but aside from Red Hat I'm not aware of many big corporations supporting projects that use it. And as you said above and as I said earlier, anything open source is good.
Comment
-
haha I know that 99% of all job offers are at least in germany only because they are cheaper than normal company advertisment so they use them as company ads.
"We are a successful company we hire all the time new people look how strong we are".
But now companies start to use them for trolling.
They could absolutly as easy write something like "we search a kernel network developer to make the linux network stack better" they would not even have to mention freebsd, but that was their goal to troll.
Its funny 99.9999% of all people on the world cannot tell if thats true what they are claiming so we have to trust em or just dont belive them. for 90% of the internets Web servers it seems that if linux is really slower it just doesnt matter. So it cant be that bad it seems.
Comment
-
Originally posted by Pawlerson View PostThis is just example from Berkeleys paper. If you really know something you wouldn't just copy paste from it. If this is just one part of the problem I'm waiting to hear about others.
As for the 3x, that was the worst case scenario model in which you'd get both read and write events each iteration. It's highly improbable, and I agree that something to the point of ~1.5 times of syscalls is the real ratio. That's still substantial if you're dealing with 10k+ active connections which this whole discussion is about. Dealing with small number of active connections can be safely and probably even more efficiently done with threads.
Thread-pooling kQueue/epoll sets is I think the best way to go tho, best of both worlds.
I'm a Linux user and consider Linux to be faster in 90% of applications but networking was always better in freeBSD at least (can't speak for other BSD flavors).
Comment
-
Originally posted by Michael_S View PostBe fair - everything that killed MySpace had to do with the user interface and the auto-play audio feature, and very little to do with Windows. Linux or FreeBSD would not have saved them from their 1997 Geocities interface.
Comment
-
Originally posted by david_lynch View PostMyspace was all about microsoft windows. I once asked someone working there why they didn't use Linux. Do you know what they said? "The things we are doing here are so cool and awesome that windows is the only possible platform for us. There's no way we could make it work with linux". Talk about living in a bubble! I think they'd gone to one too many microsoft pep rallies. Later on, one department built an ad server farm using Linux, and it was the only part of the whole operation that was making money. But it was too little, too late. I went to interview at Fox for a job with the up and coming linux team, and seriously considered their offer because of all the cool benefits, but in the end, the killer traffic in Beverly hills dissuaded me and I stayed where I was. But there is no doubt in my mind that their demise was directly tied to their blind devotion to microsoft windows.
Comment
-
Originally posted by Almindor View PostI didn't even read that paper before Michael brought it up. I worked with epoll and kQueue both and speak from experience. Subjectively it also seemed to me that the system CPU usage was lower on freeBSDs under bigger networking stress than in Linux, but I don't have the numbers to prove it.
As for the 3x, that was the worst case scenario model in which you'd get both read and write events each iteration. It's highly improbable, and I agree that something to the point of ~1.5 times of syscalls is the real ratio. That's still substantial if you're dealing with 10k+ active connections which this whole discussion is about. Dealing with small number of active connections can be safely and probably even more efficiently done with threads.
Thread-pooling kQueue/epoll sets is I think the best way to go tho, best of both worlds.
I'm a Linux user and consider Linux to be faster in 90% of applications but networking was always better in freeBSD at least (can't speak for other BSD flavors).
Comment
-
Originally posted by Pawlerson View PostIt seems you're right about this. I've found some benchmarks which show great performance of kQueue and coincide with your tests. It will be nice to have its equivalent in Linux. It's the first time I got hard evidence of someones claims at Phoronix.
Comment
-
Originally posted by Michael_S View PostBe fair - everything that killed MySpace had to do with the user interface and the auto-play audio feature, and very little to do with Windows. Linux or FreeBSD would not have saved them from their 1997 Geocities interface.
Comment
Comment