Submitted by Linux distributions come and go by the dozens almost every day, and most of them live and die an unknown, irrelevant life, mostly because no, changing three icons and adding the suffix ‘-nix’ to any random word doesn’t make it different from Ubuntu. Anyway, sometimes, a new distribution is started that brings something new to the table. One such “distribution” is Glendix, which aims to combine the Linux kernel with the userpsace tools from Plan 9. Distribution is probably not the right term for this project.
We all know what Linux is, but what exactly is Plan 9? Plan 9, developed at Bell Labs, was supposed to become the successor to UNIX, also developed at Bell Labs. Plan 9 was written from scratch, and made many radical departures from standard UNIX conventions. For instance, in Plan 9, everything really is a file; even the window a program is running in is represented as a file in a hierarchical file system. Every program in Plan 9 sees every possible resource as a file. Plan 9 is also fully distributed, so that parts of the operating system can run on different machines. Plan 9 never made its way out of the research departments, and the reason is fairly simple: UNIX is good enough. It might not be exactly elegant, but it gets the job done.
The Glendix developers aim to combine the Plan 9 userspace tools with the Linux kernel. According to them, the Plan 9 kernel is limited when it comes to hardware support, and this is where Linux comes into play.
In this project, we decouple Linux from GNU utilities, and port Plan 9 user-space applications to run on the Linux kernel. In summary, we are combining the Plan 9 user-space with the Linux kernel-space – resulting in a hybrid operating system. We think this would o er the best of both worlds – great hardware support with a cutting-edge application development environment.The primary goal of the project is to create a Linux based operating system that includes the most important user-space applications from Plan 9. For brevity, we restrict our work to only the Intel x86 architecture.
The project is obviously not done or ready for mass consumption, but you can already dive into the source code. There are also two papers you can read for more information.
The truth is, many people has been working into bringing Plan9 ideas to Linux. There’s 9P protocol support, FUSE can be used to implement userspace filesystems and there has been a lot of work into making possible per-user FS namespaces. Sure, it’s not so “pure” as Plan9, but the features are there.
IOW: it may not be worth to release a “Plan 9 Linux distro” – just work with the regular Linux distros to implement and use Plan9 features.
consider this the proving grounds for such tools on Linux.
“Plan 9 from Outer Space” was a movie made by Edward D. Wood, Jr. “Glen or Glenda” is another movie by the same director. Thus – Glendix.
Glenda, the bunny, is the Plan 9 mascot. Like Tux the penguin is for Linux:
http://en.wikipedia.org/wiki/Glenda,_the_Plan_9_Bunny
And Glenda’s supercomputing cluster at IBM:
http://tinyurl.com/2bz6mf
Edited 2008-11-26 23:10 UTC
That looks photoshopped. I can tell from the pixels, having seen quite a few in my time.
Possibly. But the port is real. More information on the port to BlueGene/L hardware is here:
http://tinyurl.com/5d83r3
Edited 2008-11-27 03:04 UTC
Ummm… No, not photoshopped. Plan 9 on BlueGene was reported on Slashdot a while ago:
http://slashdot.org/article.pl?sid=07/06/19/1215253
Plan 9 isn’t some fly-by-night OS. It was written at Bell Labs as the experimental successor to UNIX. A number of features incorporated into UNIX have come from it /proc, devfs, and the Unicode to name a few.
Obligatory wikipedia mention:
http://en.wikipedia.org/wiki/Plan_9_from_Bell_Labs
The authors are the same guys who originally took part in writing UNIX. It goes without saying that they know what they’re doing.
It doesn’t have to be Linux to run on a supercomputing cluster.
GC
It isn’t? Certainly does look shopped.
And as we know everything posted Slashdot is truth!
Yeah… I have tried Plan 9 a few years ago. They know what they are doing in system design/research, but their aim was not producing a realworld OS. Plan 9 crashes all the time (it’s also slow/not optimized and has nonexistent hardware support, but these don’t matter on a cluster).
As far as I know there is a Plan 9 cluster run by Ron Minnich and Andrej Mirtchowski at a national laboratory (and it has no bunnies stitched to the racks…), but none at IBM.
GG
What exactly were you running it on? I’ve never had a problem with Plan 9 crashing at all.
Besides… just because it crashes on your x86 doesn’t mean it won’t run smoothly elsewhere.
And, no, I didn’t say everything that is posted on slashdot is the truth, but the story does seem credible. If you search “Plan 9″+BlueGene on google you’ll see that there has been work at IBM on this.
So… you’re basing your disbelief on some pixels you think you see? Could you be specific beyond a “feeling” on what is leading you to believe that it is photoshopped? Or is it simply that you don’t believe that anything but Linux could do this?
GC
Architecturally, Plan 9 is ideal for this use. Built from the ground up as a distributed system, the concept of “processor server” is basic to it (as are display servers, storage servers, print servers, camera servers, etc.), all communicating happily via 9P. As to how well optimized it, and its compiler is for this use, I don’t know. According to the Glendix paper, KenCC, written in only 1/17th as many lines of code as gcc, can do everything that gcc can. One could be forgiven for having a few doubts about that, though.
Edited 2008-11-27 15:03 UTC
Have you ever seen GCC’s source code? I can tell you from experience that I have no problems believing that a smaller, better written compiler is possible. There’s a decent amount of bloat there.
Yeah, I’ve seen quite a few pixels in my time too. Oh hot damn, there’s another one.
</smartass>
“That reminds me of someone. You know who I mean… that person, the one who’s always standing and walking.”
If you look really carefully, under a magnifying glass, you can see that each pixel has an Adobe logo in the bottom right corner.
BTW, I did some research, and the photo was, indeed, photoshopped by “Andrey”:
http://graverobbers.blogspot.com/2007/04/more-plan-9-on-blue-gene.h…
Not as a hoax, but as an upfront little joke.
Plan 9 on BG/L is real. And I would be very interested in information regarding how useful it is in that capacity.
Edited 2008-11-27 19:14 UTC
Yes, and even just at a casual glance, the logos appear way too uniform to be anything other than a copy-paste job.
It was the wording of the original “This looks photoshopped because of the pixels” post that I found amusing, not the notion itself.
Why you know this is beyond me.
Linux reviews come and go by the dozens almost every day, and most of them live and die an unknown, irrelevant life, mostly because no, changing three words and adding the prefix “new and improved” to any old name doesn’t make today’s review different from the last review.
That said, welcome to Glendix! The more the merrier, the website looks lovely and besides, I like the mascot.
Edited 2008-11-27 00:06 UTC
Distribution is still spot on.
the power and glory of UTF-8, and BeOS brought it to the desktop.
Hallowed is Plan9!
As as the long default WM isn’t 9wm; it has to be the most cumbersome WM ever devised. Ugly. Your cursor is taken over until you manually draw a new window on the screen. No. Just no.
Section 2.1 of the comprehensive project report strains credibility a bit:
http://glendix.org/report.pdf
I’m not a huge fan of GNU, and always like to see alternatives. But I think I’d like to see some proof of the assertions made.
I would retract that rather overzealous statement if I could! However, the fact remains that GCC *is* rather bloated, even if the Plan 9 C compilers don’t offer all the features GCC does.
A drop in the bucket compared to the overzealous statements I’ve heard from GNU software fans. 😉
Anyway, since I have a Plan9 knowledgeable person handy, and haven’t been able to find it anywhere else, may I ask what are the current limitations of Fossil? Max file size? Max volume size? 9P gets all the press, but it’s really just a wire protocol, like NFS, and not really a filesystem. (Not to belittle 9P. In an OS where *everything* is truly a file, the filesystem wire protocol takes on a whole new significance.)
Also, any comments on where Plan9 or Plan9 ideas have proved useful in the real world. UTF-8… (and powering the bell-labs.com site), of course. Other stuff?
Edited 2008-11-27 19:29 UTC
To be honest I have not found much information about the max file size or the max volume size of Fossil/Venti (Venti is the actual storage server). I did however find a paper referencing Fossil/Venti’s abilitiy to hold at least an exabyte of data. I would assume Plan9 has a large storage capacity but the information is slim.
…Minix in facts ? It sounds like it is a micro-kernel mockup, yet with the Linux spices to brand all of this with the -Nix suffix. Hey, waddaminute, Minix DO ends with -Nix also ! So what’s really new ?
Kochise
Btw, is it me or the Raving Rabbits look suspiciously Glenda like ?
…..both without the GNU cruft.
AnubisOS and now Glendix, who cares about the state of the global economy? Exciting times are ahead!
http://anubis-os.org http://arix.sf.net
Lines like this always remind me of an idiotic remark someone made about Tron after Matrix came out, i.e. that it was the Matrix of the eighties. I don’t really care what people compare to Ubuntu or what they compare Ubuntu with, I just never got convinced that Ubuntu is the best there is, thus stating something isn’t far from Ubuntu won’t place either it below or above it.
he wasn’t talking about that, he was talking about the eight billion Ubuntu “derivative” distributions.
Not to mention Debian derivatives?
I never hear Debian mentioned anymore.
Debian is implicit in any mention of the Ubuntu family. In my opinion, Debian’s strength and destiny lies with providing a solid foundation for derivative distros. I view Debian Stable as a sort of reference implementation to keep them on track.
Implicit to your or me, but Ubuntu is pulling in all kinds of people who don’t know anything about Linux, or the history of Ubuntu.
I ran Slack and Debian before Ubuntu was born… both were, and still ARE excellent distributions… but I understand that the success of marketing Ubuntu is also (in a way) success for linux in general.
I know Debian still lives and breaths, I just feel like it is never mentioned anymore… and that makes me sad because it is an excellent package.
http://www.google.com/search?hl=en&q=site%3Awww.ubuntu.com+debi…
Edited 2008-11-28 19:12 UTC
O.K. You win. Debian is everywhere and is mentioned frequently. There are a lot of Debian articles on OSNews and slashdot and everywhere else too. In fact, now that you mention it, I just turned around and saw the debian symbol in the stucco on my wall.
So based on that the GNU has a lot of more lines they should use the plan 9 tools instead?
Anyone who doesn’t believe that IBM really has a BG/L Plan 9 Cluster is clearly uninformed. There are several papers and presentations online that describe the machine specifices and the hurdles encountered when building the machine. They also answer the question “Why not Linux?”. The short answer is that Linux has a much longer, slower I/O path than Plan 9.
Here’s the original OSNews article:
http://osnews.com/comments/18102
Edited 2008-11-27 16:22 UTC