It's not a very useful benchmark, sorry to say. They might be loading on all evenly, but is the necessary kernel scheduling to keep them all fair being done efficiently? You might be losing a lot of cycles that way. Similarly, device drivers dealing with interrupts need to be scheduled sensibly.
What you could do is dual-boot with Linux on the same machine (heck, even using a Gentoo LiveCD with an SMP kernel) and run an Apache benchmark or something, at least as long as it uses SMP. Post the results here. While I doubt anybody expects NetBSD to beat Linux, it's worth it to know where it stands.
If for nothing else (and I really am shaving off a lot of value), NetBSD is a top-notch educational tool. Clean code is easy to learn from. Where university courses examine Linux for how a kernel should work, they should be examining NetBSD. You can argue all you want about this but it remains true, it's infinitely cleaner and hence useful to learn from.
As an operating system itself, it does continue to be very useful. How is it redundant? It has many of the advantages of other systems without their disadvantages. Was it you or someone else who replied to my previous mention of this with "doesn't make a good OS"? Of course it does. Strengths of others without their weakness makes the PERFECT OS. While NetBSD is not perfect - there are many aspects missing that could bring it to desktop use - it comes much closer than other projects I've seen. Running on plenty of architectures usefully is just another great advantage in this sense, that even if x86 dies (and this is happening, in favor of AMD64) the system will still be strong. Look at FreeBSD 5: apart from x86 it isn't 'complete' anywhere, bits of the new functionality still aren't ported. Linux 2.6 broke many ports (MIPS, for instance) that still aren't completely fixed yet, too (unless you can show me an SGI Indy running Linux 2.6...). NetBSD 2 jumped forward with the ports intact. That's portable, and it won't die out.
I recall seeing posts here complaining about Slashdot being a month or a year behind the news, but almost a century? That's gold. They must be really desperate for news.
The ISOs were up, but the release itself was not announceed, and releases are only posted once they are announced... makes sense since the post is invariably just a trim of the release announcement. I fetched the ISOs two days ago, and yesterday the release was announced (Australia days) and posted. Interestingly it got more posts in that one day than the last few BSD threads combined.
There are easier ways than XBox. It's 'fun' to have the XBox, yes, but when you consider that NetBSD on anything that boots is already a fully functional serving or development environment, you start to find cheaper ways. Hunt around universities or workplaces for when they throw away old junk, ask if you can take it "for your sick grandfather who just wants to say he once had a computer", install NetBSD in whatever way necessary, and give this to the less fortunate students.
Personally I picked up an Indy (with a muscular R5000) in a similar way, but it was more through a friend of a friend who hunted around unis. The machine cost $5k back in its day, now it's free, and with a larger hard drive than what I ended up with, would be a really useful machine.
As for XBox support, I don't know why it's not supported yet; I've read around and found that someone was making healthy progress but got stuck on figuring out the "right" way to do things (NetBSD/xbox or NetBSD/i386? New drivers or just quirks?), and haven't heard anything about the project since.
Re:Printed documentation (diff NET/FREE BSD)
on
NetBSD 2.0 Released
·
· Score: 4, Informative
You can print the NetBSD handbook yourself:)
There are many similarities between FreeBSD and NetBSD thanks to their mutual heritage, but FreeBSD's documentation doesn't usually apply equally to NetBSD. The differences are well covered in NetBSD's own online documentation, though.
I had been using FreeBSD since 4.8 or so, and was able to pick up NetBSD almost instantly. Only one thing held me back (for weeks even), and that was my use of CFLAGS= instead of CFLAGS+= in mk.conf, which made world builds break. Entirely my fault, but could use a warning in documentation. But the basic idea is, if you're willing to read a couple of items of documentation and ask questions, it's very easy to learn.
Re:They are quite clear on this.
on
NetBSD 2.0 Released
·
· Score: 2, Informative
That's FUD. You would be more right in saying 90% are very stable and 10% less so, but I would say the share is even more towards stable. If you look at the ports page and the mailing lists, almost all systems (save those that nobody expects much from, e.g. Playstation 2) have numerous users with success stories, and where there are problems, solutions come up.
I personally have seen many, many reports of NetBSD on exotic machines being very useful and stable. Googling is the least amount of work needed to find more.
Shame really, any real BSD follower (especially FreeBSD where it's arguably the easiest to upgrade, at least given the track record of virtually no failed compiles in -STABLE) would update his/her -STABLE very often, at least to deal with exploits... running 4.6 in this day and age, although stable, is not a good idea for a publicly available server.
I won't say much good about Apache either, which in its overcomplicates (yes you heard me) code has bred quite a few exploits. I'm just one of those guys writing his own microscopic web server (already does HTTPS) to save at least a few people from Apache's bloat.
Re:What are NetBSD's strengths?
on
NetBSD 2.0 Released
·
· Score: 2, Interesting
OpenSSH, PF, the TCP/IP stack that founded the internet... you're right, nothing out of BSD.
If anything, nothing comes out of Linux. BSDs are breeding grounds for world-changing software. Unless you mean to tell me that Linus and his buddies write all the software instead of getting it from GNU and other devs, GNU/Linux is much more of a hand-me-down collection than any given BSD, the latter containing some source that started in BSD and continues to be in BSD. Even some GNU tools (indent, for instance) were forks of BSD tools.
Just rip out the fans, and unplug it. NetBSD 2 has support for running without power or heat. It can also use your machine's case as a makeshift hard disk, if you use a PCI Pencil device (also available in USB and SBUS).
I hope you people realise this is just a joke based on NetBSD's ability to milk great value out of un-great hardware.
It DOES depend on definition. I could call efficiency code efficiency, in which case, yes, it is very efficient; BGL systems have simple code.
Apparently on a 2-way SMP system there's no real performance difference between Linux and NetBSD 2 (this is just what I've heard; don't ask ME for the numbers, Google it), same kind of story on hyperthreading (actually, apparently NetBSD gets faster with HTT on, and Linux gets slower; strange). I haven't heard of anything about 4-way SMP or anything, but no doubt the Linux rigs would start getting faster and faster because of the more scalable SMP algorithms
I'd rather see DragonFly's LWKT model on NetBSD than a BGL, because it really does have plenty of advantages and is a clean design. Who knows, maybe they'll get around to it in 3.0?
Nobody would complain. OpenBSD has a more interesting community spirit; not happy just being technically excellent, they have to be culturally integrated as well. And thanks to Theo's work, software-politically too.
NetBSD seems less interested in that kind of thing, focusing more on the code. If anyone uses it for hype-inspiring projects (like the internet speed records), it's a help, but it's their choice.
You could have gone one better and fetched netbsd-2-0 tagged CVS source, and built it locally WITH the code optimizations and post-release improvements that a source build offers.
Who else thinks that, for such a gloriously large and powerful OS, a 200MiB ISO is just amazing? Well, all the BSDs have very small install ISOs (at least, if you compare with FreeBSD's "minimal install", not the with-packages ISO), really.
You're unlucky. I built an entire system for sgimips on an i386 (well, a Pentium 3, but you get the idea) with a one-liner build.sh, and there wasn't a single problem. It didn't require any external software or version mangling, it all just worked from the toolchain in the source tree. Same for the kernel.
You can just set your fastest machine on the network, regardless of architecture, to compile distributions for all the other systems and install over NFS.
People forget that portability isn't just about booting on architectures. NetBSD provides an operating system everywhere it goes; and it provides all the tools in-tree to build and serve for these systems, even diskless installs or even roots. Is there even a single Linux distribution that can do this? And yes, it does have to be all in-tree, no redundancy or external software.
It's not on by default because it's too new a feature. BSDs work on 'method of least surprise'. If you uncomment that and build a fresh kernel, it will use the Shift+PgUp/Down mechanism that Linux has, no worse.
Why is it that if someone said Linux was much faster, you'd believe it without a number? You're very selective in your scrutiny. I bet you've also never tried NetBSD either.
I can't give you numbers because NetBSD has replaced Linux on my machines, and so Linux won't see any of them again. But I have vivid memories of it taking Linux a lot longer to do the same things, especially where bonnie had a say in things. Why does it take upward of 6 hours to build a full Gentoo base system, yet a NetBSD base system takes 2 hours on the same machine with the same compiler and same flags and produces MORE software? (Much more; several servers, gdb, and so on, of which Gentoo has none). That's on a P3 1Ghz by the way. If you want numbers, go make them yourself. You've heard my anecdote.
Oh, and I don't want to hear people using GNU software bloat (which certainly helps the compile time take ages) as an excuse for the long compile time - you're helping *BSD there.
Because you want a system that's engineered, not just hacked until it compiles? People who run BSD usually do so because they want systems that are Right, not because of some little performance difference. And as a personal anecdote NetBSD hugely outperforms Linux 2.6 on all of the machines I have tried in my experience, including an SGI Indy. While Linux performs "reasonably close to hardware speed under normal conditions", NetBSD can perform at (sometimes above, thanks to the clever UVM) hardware speed under any conditions. It sounds silly but you really have to try it.
Re:Its good at nothing, and ignored for a reason.
on
NetBSD 2.0 Released
·
· Score: 1
It has many advantages over the others, they just aren't advertised much. Similarly it's also very balanced - modern and portable like Linux, faster than FreeBSD 5, security record almost as good as OpenBSD (in practice, just as good), and cleaner than all of them in every way that counts. NetBSD could be THE BSD since it appeals to more needs on more architectures than the others; its only shortcoming, really, being lack of corporate support (I wouldn't run Linux at all if there was a working NVidia driver for NetBSD).
Re:Ah. Blissful clean architecture.
on
NetBSD 2.0 Released
·
· Score: 4, Informative
The benchmarks are a year old, the system used is even older. Anyway, what's your point? They didn't bother with scalability until recently. You'd be amazed at what NetBSD 2.0 can do. Go try it yourself. Condemning an OS based on not being scalable at one point in time is just stupid. Linux wasn't scalable until 2.6, have you condemned that too? "Look at these benchmarks from 2 years ago - it shows a very interesting picture of Linux sucking".
On a related note, it isn't just NETWORK socket performance, since you can use sockets over loopback too. In NetBSD, being so supportive of systems which need as much space as possible, can even compile a replacement pipe mechanism which uses sockets to be smaller but slightly slower.
Really? I love the new installer and I'm not alone in this. It's actually good to have a logical and simple installer that still does everything an installer should. What, you'd rather do Gentoo? Please.
1: I explained this in another post, you must have missed it. The BSDs can have sources fetched via CVS (NetBSD recommends this way, rightly so), and having it in the base package makes this infinitely more convenient than having to install the gargantuan cvsup port or poking around for up-to-date-enough source tarballs once daily. Given the relatively small footprint of the CVS client, this convenience is well worth it.
2: They don't have 'too much human resources', you're thinking of Linux. OpenBSD has clear goals and, yes, are motivated to achieve these goals. Security and freedom are goals; this project helps both. The BSDs don't "struggle hard" with manpower, they have as many developers as are needed; everything worth doing gets done. And having less developers is often better for coordination, which is why BSD code bases continue to be consistent and robust.
Slashdot Mirror
It's not a very useful benchmark, sorry to say. They might be loading on all evenly, but is the necessary kernel scheduling to keep them all fair being done efficiently? You might be losing a lot of cycles that way. Similarly, device drivers dealing with interrupts need to be scheduled sensibly.
What you could do is dual-boot with Linux on the same machine (heck, even using a Gentoo LiveCD with an SMP kernel) and run an Apache benchmark or something, at least as long as it uses SMP. Post the results here. While I doubt anybody expects NetBSD to beat Linux, it's worth it to know where it stands.
It won't be worse than Slackware. And yes, it functions as a desktop. Just try it and see. Be prepared to read the documentation on the site, though.
If for nothing else (and I really am shaving off a lot of value), NetBSD is a top-notch educational tool. Clean code is easy to learn from. Where university courses examine Linux for how a kernel should work, they should be examining NetBSD. You can argue all you want about this but it remains true, it's infinitely cleaner and hence useful to learn from.
As an operating system itself, it does continue to be very useful. How is it redundant? It has many of the advantages of other systems without their disadvantages. Was it you or someone else who replied to my previous mention of this with "doesn't make a good OS"? Of course it does. Strengths of others without their weakness makes the PERFECT OS. While NetBSD is not perfect - there are many aspects missing that could bring it to desktop use - it comes much closer than other projects I've seen. Running on plenty of architectures usefully is just another great advantage in this sense, that even if x86 dies (and this is happening, in favor of AMD64) the system will still be strong. Look at FreeBSD 5: apart from x86 it isn't 'complete' anywhere, bits of the new functionality still aren't ported. Linux 2.6 broke many ports (MIPS, for instance) that still aren't completely fixed yet, too (unless you can show me an SGI Indy running Linux 2.6...). NetBSD 2 jumped forward with the ports intact. That's portable, and it won't die out.
I recall seeing posts here complaining about Slashdot being a month or a year behind the news, but almost a century? That's gold. They must be really desperate for news.
This just in: I brushed my teeth.
The ISOs were up, but the release itself was not announceed, and releases are only posted once they are announced... makes sense since the post is invariably just a trim of the release announcement. I fetched the ISOs two days ago, and yesterday the release was announced (Australia days) and posted. Interestingly it got more posts in that one day than the last few BSD threads combined.
There are easier ways than XBox. It's 'fun' to have the XBox, yes, but when you consider that NetBSD on anything that boots is already a fully functional serving or development environment, you start to find cheaper ways. Hunt around universities or workplaces for when they throw away old junk, ask if you can take it "for your sick grandfather who just wants to say he once had a computer", install NetBSD in whatever way necessary, and give this to the less fortunate students.
Personally I picked up an Indy (with a muscular R5000) in a similar way, but it was more through a friend of a friend who hunted around unis. The machine cost $5k back in its day, now it's free, and with a larger hard drive than what I ended up with, would be a really useful machine.
As for XBox support, I don't know why it's not supported yet; I've read around and found that someone was making healthy progress but got stuck on figuring out the "right" way to do things (NetBSD/xbox or NetBSD/i386? New drivers or just quirks?), and haven't heard anything about the project since.
You can print the NetBSD handbook yourself :)
There are many similarities between FreeBSD and NetBSD thanks to their mutual heritage, but FreeBSD's documentation doesn't usually apply equally to NetBSD. The differences are well covered in NetBSD's own online documentation, though.
I had been using FreeBSD since 4.8 or so, and was able to pick up NetBSD almost instantly. Only one thing held me back (for weeks even), and that was my use of CFLAGS= instead of CFLAGS+= in mk.conf, which made world builds break. Entirely my fault, but could use a warning in documentation. But the basic idea is, if you're willing to read a couple of items of documentation and ask questions, it's very easy to learn.
That's FUD. You would be more right in saying 90% are very stable and 10% less so, but I would say the share is even more towards stable. If you look at the ports page and the mailing lists, almost all systems (save those that nobody expects much from, e.g. Playstation 2) have numerous users with success stories, and where there are problems, solutions come up.
I personally have seen many, many reports of NetBSD on exotic machines being very useful and stable. Googling is the least amount of work needed to find more.
Shame really, any real BSD follower (especially FreeBSD where it's arguably the easiest to upgrade, at least given the track record of virtually no failed compiles in -STABLE) would update his/her -STABLE very often, at least to deal with exploits... running 4.6 in this day and age, although stable, is not a good idea for a publicly available server.
I won't say much good about Apache either, which in its overcomplicates (yes you heard me) code has bred quite a few exploits. I'm just one of those guys writing his own microscopic web server (already does HTTPS) to save at least a few people from Apache's bloat.
OpenSSH, PF, the TCP/IP stack that founded the internet... you're right, nothing out of BSD.
If anything, nothing comes out of Linux. BSDs are breeding grounds for world-changing software. Unless you mean to tell me that Linus and his buddies write all the software instead of getting it from GNU and other devs, GNU/Linux is much more of a hand-me-down collection than any given BSD, the latter containing some source that started in BSD and continues to be in BSD. Even some GNU tools (indent, for instance) were forks of BSD tools.
Just rip out the fans, and unplug it. NetBSD 2 has support for running without power or heat. It can also use your machine's case as a makeshift hard disk, if you use a PCI Pencil device (also available in USB and SBUS).
I hope you people realise this is just a joke based on NetBSD's ability to milk great value out of un-great hardware.
It DOES depend on definition. I could call efficiency code efficiency, in which case, yes, it is very efficient; BGL systems have simple code.
Apparently on a 2-way SMP system there's no real performance difference between Linux and NetBSD 2 (this is just what I've heard; don't ask ME for the numbers, Google it), same kind of story on hyperthreading (actually, apparently NetBSD gets faster with HTT on, and Linux gets slower; strange). I haven't heard of anything about 4-way SMP or anything, but no doubt the Linux rigs would start getting faster and faster because of the more scalable SMP algorithms
I'd rather see DragonFly's LWKT model on NetBSD than a BGL, because it really does have plenty of advantages and is a clean design. Who knows, maybe they'll get around to it in 3.0?
Sing one, then :)
Nobody would complain. OpenBSD has a more interesting community spirit; not happy just being technically excellent, they have to be culturally integrated as well. And thanks to Theo's work, software-politically too.
NetBSD seems less interested in that kind of thing, focusing more on the code. If anyone uses it for hype-inspiring projects (like the internet speed records), it's a help, but it's their choice.
You could have gone one better and fetched netbsd-2-0 tagged CVS source, and built it locally WITH the code optimizations and post-release improvements that a source build offers.
Who else thinks that, for such a gloriously large and powerful OS, a 200MiB ISO is just amazing? Well, all the BSDs have very small install ISOs (at least, if you compare with FreeBSD's "minimal install", not the with-packages ISO), really.
You're unlucky. I built an entire system for sgimips on an i386 (well, a Pentium 3, but you get the idea) with a one-liner build.sh, and there wasn't a single problem. It didn't require any external software or version mangling, it all just worked from the toolchain in the source tree. Same for the kernel.
You can just set your fastest machine on the network, regardless of architecture, to compile distributions for all the other systems and install over NFS.
People forget that portability isn't just about booting on architectures. NetBSD provides an operating system everywhere it goes; and it provides all the tools in-tree to build and serve for these systems, even diskless installs or even roots. Is there even a single Linux distribution that can do this? And yes, it does have to be all in-tree, no redundancy or external software.
You must be doing something really exotic.
[8:39:51pm] root@dirk conf% uptime
8:39PM up 10:51, 2 users, load averages: 0.22, 0.23, 0.24
NetBSD 2.0 system installed yesterday, booted 11 hours ago. (Yes, I installed it yesterday, the ISOs were up already).
I think I heard about it somewhere else on /.
Maybe poke around on Google? Failing that ask on a mailing list.
Or, the obvious, just run NetBSD.
This, folks, is an example of a clueless newbie insulting an operating system based on his own ignorance.
/usr/src/sys/arch/i386/conf/GENERIC:
# console scrolling support.
#options WSDISPLAY_SCROLLSUPPORT
It's not on by default because it's too new a feature. BSDs work on 'method of least surprise'. If you uncomment that and build a fresh kernel, it will use the Shift+PgUp/Down mechanism that Linux has, no worse.
Anything else you want to be owned on?
Why is it that if someone said Linux was much faster, you'd believe it without a number? You're very selective in your scrutiny. I bet you've also never tried NetBSD either.
I can't give you numbers because NetBSD has replaced Linux on my machines, and so Linux won't see any of them again. But I have vivid memories of it taking Linux a lot longer to do the same things, especially where bonnie had a say in things. Why does it take upward of 6 hours to build a full Gentoo base system, yet a NetBSD base system takes 2 hours on the same machine with the same compiler and same flags and produces MORE software? (Much more; several servers, gdb, and so on, of which Gentoo has none). That's on a P3 1Ghz by the way. If you want numbers, go make them yourself. You've heard my anecdote.
Oh, and I don't want to hear people using GNU software bloat (which certainly helps the compile time take ages) as an excuse for the long compile time - you're helping *BSD there.
Because you want a system that's engineered, not just hacked until it compiles? People who run BSD usually do so because they want systems that are Right, not because of some little performance difference. And as a personal anecdote NetBSD hugely outperforms Linux 2.6 on all of the machines I have tried in my experience, including an SGI Indy. While Linux performs "reasonably close to hardware speed under normal conditions", NetBSD can perform at (sometimes above, thanks to the clever UVM) hardware speed under any conditions. It sounds silly but you really have to try it.
It has many advantages over the others, they just aren't advertised much. Similarly it's also very balanced - modern and portable like Linux, faster than FreeBSD 5, security record almost as good as OpenBSD (in practice, just as good), and cleaner than all of them in every way that counts. NetBSD could be THE BSD since it appeals to more needs on more architectures than the others; its only shortcoming, really, being lack of corporate support (I wouldn't run Linux at all if there was a working NVidia driver for NetBSD).
The benchmarks are a year old, the system used is even older. Anyway, what's your point? They didn't bother with scalability until recently. You'd be amazed at what NetBSD 2.0 can do. Go try it yourself. Condemning an OS based on not being scalable at one point in time is just stupid. Linux wasn't scalable until 2.6, have you condemned that too? "Look at these benchmarks from 2 years ago - it shows a very interesting picture of Linux sucking".
On a related note, it isn't just NETWORK socket performance, since you can use sockets over loopback too. In NetBSD, being so supportive of systems which need as much space as possible, can even compile a replacement pipe mechanism which uses sockets to be smaller but slightly slower.
Or, you know, cvs + ./build.sh :)
Really? I love the new installer and I'm not alone in this. It's actually good to have a logical and simple installer that still does everything an installer should. What, you'd rather do Gentoo? Please.
1: I explained this in another post, you must have missed it. The BSDs can have sources fetched via CVS (NetBSD recommends this way, rightly so), and having it in the base package makes this infinitely more convenient than having to install the gargantuan cvsup port or poking around for up-to-date-enough source tarballs once daily. Given the relatively small footprint of the CVS client, this convenience is well worth it.
2: They don't have 'too much human resources', you're thinking of Linux. OpenBSD has clear goals and, yes, are motivated to achieve these goals. Security and freedom are goals; this project helps both. The BSDs don't "struggle hard" with manpower, they have as many developers as are needed; everything worth doing gets done. And having less developers is often better for coordination, which is why BSD code bases continue to be consistent and robust.