Slashdot Mirror

If you look at the GENERIC kernel config file for 6.0 Release, it shows the ULE scheduler is not enabled by default, the 4BSD scheduler still is. http://www.freebsd.org/cgi/cvsweb.cgi/~checkout~/s rc/sys/i386/conf/GENERIC?rev=1.429.2.3.2.1&content -type=text/plain&only_with_tag=RELENG_6_0_0_RELEAS E The basic benefit the ULE scheduler was designed to have is O(1) performace for the SMP case. Here's the BSDcon '03 whitepaper Jeff Robertson wrote about ULE http://www.chesapeake.net/~jroberson/ULE.pdf Yes it's been fixed up a lot, you can read kerneltrap's coverage of it a few times, but you'd have to dig into the mailing lists to see why it is not default still, which I didn't do.

Although ULE may be 'more' fixed in FreeBSD 5.4, I still find it to be very unstable. I'm unable to have a system booted for more than a few minutes before it kernel panics. (I'm also ironically unable to recompile a GENERIC kernel all the way through).

I was very excited to try the ULE Scheduler for FreeBSD after having read the excellent article: ULE: A Modern Scheduler for FreeBSD.

I eagerly await the next version of FreeBSD, in hopes that ULE is mature enough to take advantage of the performances increases of this more modern Scheduler.
This may however prove to be a moot point in light of the recent P4 HTT insecurities that have been revealed.

For the record, FreeBSD's scheduler was already O(1), it just didn't handle SMP extremely well. The new ULE scheduler handles the SMP case much better, along with other nice improvements. See Jeff's paper at http://www.chesapeake.net/~jroberson/ULE.pdf

The release engineering team certainly does have high standards. Trying to live up to the stability reputation. But keep in mind that 5.x still is considered in testing and major changes can still be afoot that can cause instabilities. So please still keep in mind what -current means, http://www.freebsd.org/doc/en_US.ISO8859-1/books/h andbook/current-stable.html, and read the early adopters guide.

5.x will get better and better as it approaches 5.3R, so while some of the problems running a -current release are lessened, one should still be aware of all this and the higher standard for fixing one's own problems when running 5.x. RTFM is not an insult when running 5.x, its simply a price of entry to a great OS.

Keep in mind that ULE is derived from Linux 2.6's scheduler. Take a look at the last sentence of Section 2 of the BSD white paper on UL.

Wrong? What do you mean wrong? Yes, Francesco Sportolari's patch ports interactivity work from FreeBSD to Linux.

The actual guts of the O(1) scheduler were written by Ingo Molnar for Linux, and that was then taken for SCHED_ULE. Read THIS

And in case you don't, I'll quote:
"Linux has also recently recieved a totally rewritten event driven scheduler from which ULE's queueing system was derived."

I didn't say that FreeBSD's 4BSD scheduler wasn't O(1), or that ULE was the default scheduler, so I don't know how that is relevant to the argument.

I won't go off making false accusations because I do have my facts straight. Get yours straight before you falsely accuse me of making false accusations.

I never said FreeBSD was bad, its great. It really gives me the shits when people try to say things like linux takes heaps of good FreeBSD code, and FreeBSD wouldn't touch Linux code even if they could type of things. Or when they say
FreeBSD is more stable
FreeBSD is more secure
FreeBSD is more scalable
FreeBSD is more robust
FreeBSD has a better networking stack
FreeBSD is better under load
etc.
Without *anything* to back them up. I mean, what the fuck do I look like? An idiot manager type who laps up the buzzwords and lack of content?!? Quite apart from half the claims being blatantly false, the other half of them have not been measured for a long time.

There are very few reasons, from a technical perspective, to use proprietary operating systems instead of GNU. Especially with the new Linux 2.6 kernel

I've met the fella that heads up the The Cross Creek Initiative. They are basically a non-profit organization promoting alternative fuels. Hydrogen is what they believe in the most. I've read some of the company reports on hydrogen they have, along with many many letters of support from various universities and scientists world-wide.

Worth checking out if you're interested in options to centralized power/alternative fuels. The fella I mentioned put it to me this way:

20 years ago when you went into a bank and asked to see a computer, they took you down to the basement and showed you the huge mainframe sitting there. At desks were dumb terminals connected to a centralized computer. Now when you go into a bank, you see 10's to 100's of seperate computers. Decentralized. Hydrogen can do the same with powering our houses, vehicles, etc....

searcher