fefe.de benchmark discussion (Linux vs. *BSD)

[n0cmonkey]

fefe.de

 

[n0cmonkey]

Originally posted by: Nothinman

Do we need to go into how these benchmarks were developed in Linux and "ported" to the other OSes?

So? Most of them didn't require any porting because they were basic C semantics with open(), read(), mmap(), fork(), etc, the only thing I remember that had OS-specific parts was the poll/notification part of the web server.

But it was still developed with a bias towards linux. Maybe it isn't a huge deal, but the other points are.

(This was posted in another thread, and I am answering it here)

 

[n0cmonkey]

Originally posted by: chsh1ca

Originally posted by: n0cmonkey
I did. Notice:

Remarks

Slashdot accepted the story, and this site is being slashdotted as of Oct 19 2003. Here are a few answers to questions that have been raised in the Slashdot forums:

So, the remarks area was UPDATED as the days went by. When I read EVERYTHING on that page, not all of those remarks were there. Stilll, it doesn't mention time of download or server used (unless he added something else after I read the page 😉).

Yes I know, but I figured rechecking the page for updates wasn't out of the realm of possibility, especially since this is an ongoing topic. 🙂

I had read several of the updates, but missed those.

Simply saying that his results are flawed because of OBSD's results doesn't prove anything either.

I admit that the performance of OpenBSD is more than lacking compared to other systems. The results of the tests on OpenBSD are not exactly what I think was wrong with the tests.

Obviously something was wrong, and to be honest, using a dev kernel probably wasn't the best idea, but I know why he did (he'd have been crucified had he not). Personally, I would like to know why exactly you think that the sum of the benchmarks are flawed. OpenBSD aside, since that seems to be your initial gripe, everything else makes linux look pretty good. I'm not sure if this is a slant, or if those are real numbers, but at least I can verify them by looking at the source of the app and running it (once I'm home). His updates have definitely improved things, and answered a few questions.

I think Jose Nazario's response to Felix's e-mail about retesting sums up the situation pretty well. Do we need to go into how these benchmarks were developed in Linux and "ported" to the other OSes?

Should we go ahead and make a new thread on this? 😛

(this was originally posted in the other thread, so I'm posting it here too)

 

[Nothinman]

But it was still developed with a bias towards linux. Maybe it isn't a huge deal, but the other points are.

I still don't see how it's biased, sure it seems like the guy knows more about Linux than the rest of the OSes but IMO the test still showed favorable towards NetBSD (especially for their quick turnaround in fixing several of the problems) and FreeBSD and I think it was as fair as any benchmark can get.

The source to all his tests are available via CVS too, if you (or someone on the OpenBSD team) think they can find a problem in the code that shows it was designed for Linux and engineered in a way that would make all those other OSes look bad I'm sure we'd all be interested and frankly I think he would be too because I don't think he tried to intentially be malicious to any of the OSes. I think the OpenBSD people just got their egos bruised because their buffer caching system basically sucks because it's not unified yet and it apparently hurts performance in some (maybe corner) cases. For what most people use OpenBSD for it's obviously not been an issue before and probably still won't be a big one atleast from a user's standpoint, unless people keep doing this kind of benchmark on it.

 

[n0cmonkey]

Originally posted by: Nothinman

But it was still developed with a bias towards linux. Maybe it isn't a huge deal, but the other points are.

I still don't see how it's biased, sure it seems like the guy knows more about Linux than the rest of the OSes but IMO the test still showed favorable towards NetBSD (especially for their quick turnaround in fixing several of the problems) and FreeBSD and I think it was as fair as any benchmark can get.

The source to all his tests are available via CVS too, if you (or someone on the OpenBSD team) think they can find a problem in the code that shows it was designed for Linux and engineered in a way that would make all those other OSes look bad I'm sure we'd all be interested and frankly I think he would be too because I don't think he tried to intentially be malicious to any of the OSes. I think the OpenBSD people just got their egos bruised because their buffer caching system basically sucks because it's not unified yet and it apparently hurts performance in some (maybe corner) cases. For what most people use OpenBSD for it's obviously not been an issue before and probably still won't be a big one atleast from a user's standpoint, unless people keep doing this kind of benchmark on it.

According to the page, the flamers were not OpenBSD developers, but users. What many people in the OpenBSD world complained about was the testing procedure. The page would be fine for someone's home page or other mostly meaningless site on the net. But this was a presentation at a Linux conference. I think the points made in the email I linked are very valid. Hell, a proper testing procedure could produce the same results. But until someone goes through a better procedure we will never know.

 

[Nothinman]

According to the page, the flamers were not OpenBSD developers, but users

The developers don't use the software and the user's don't develop the software? I thought this was open source ;p

 

[n0cmonkey]

Originally posted by: Nothinman

According to the page, the flamers were not OpenBSD developers, but users

The developers don't use the software and the user's don't develop the software? I thought this was open source ;p

Ok, fine. 😛

Here is the quote:

[Update Oct 22 2003]: While I received a lot of verbal abuse from OpenBSD users, the OpenBSD project itself is focusing on solving the issues; the 1024 cylinder limit is apparently being worked on, and the fork panic has been reproduced and a bug has been opened in the bug tracking system.

I took this to mean the usual developers were not to blame. 😉

 

[chsh1ca]

Originally posted by: n0cmonkey
I think Jose Nazario's response to Felix's e-mail about retesting sums up the situation pretty well. Do we need to go into how these benchmarks were developed in Linux and "ported" to the other OSes?

Should we go ahead and make a new thread on this? 😛

(this was originally posted in the other thread, so I'm posting it here too)

Okay, well, regardless of how the benchmarks were developed, they do, as indicated, use the different native calls and event solutions provided by each OS. The reality is that the benchmarks may have indeed been much more optimized for each architecture than the Apache webserver. Keeping in mind the whole goal of this set of benchmarks was to see which OS was the best scalable webserver. IMO, a better range of tests would be dedicated, identically configured boxes running the software in question (Apache) trying to serve out hundreds to thousands of simultaneous requests. Such tests are not too difficult to do, however, they are time consuming, and you'd need more substantial amounts of hardware than one laptop. 🙂 I would favour such tests over a small app designed to test the performance of the standard system calls that are made, since there are other factors that should be taken into account.

That being said, if the code is ported properly, then it should be a non-issue, since during the porting the appropriate calls would be modified. The peer review idea might have been a good suggestion before the benchmarks were released, but I noticed he had utterly nothing to say about the code, which presumably he read. If indeed there were an issue with how the benchmark is written, why not bring it up? It seems to me that a lot of OBSD people are just sore it did so poorly. If the benchmarks are accurate, it becomes painfully evident just HOW MUCH performance the OBSD kernel devs have been giving up in favour of focusing on security. Nobody sane is going to think OBSD is a bad OS because of this benchmark, since it is a very specific application, and doesn't deal with the other administrative issues (security, updates, et. al.). So it may not be the best of the free nixes for a high performance scalable webserver (that seems to be Linux 2.6). It certainly has one of the best security track records in the industry, something nobody else can claim.

 

[drag]

Well...

I agree that actual webserver performance using Apache would be usefull, it may not practicle for 2 reasons.

1. It would be just to darn expensive. He would have to have a dedicated webserver and a bunch of high-powered cllient machines to do it. (as you pointed out.) And probably would of had to had a couple other guys helping him.

2.(and the bigger deal) People would freak out on him 10x as much as now. The version of Apache he used, php scripts, perl scripts, how big the files were. What network card he used, what drivers he used, etc etc. Every little detail, every little thing that could give one OS a artificial advantage over the other one. His e-mail box would get hammered.

I think how he did it would probably the best one guy could do in a resonable time alloted.

Now on a side note.

I like OpenBSD and all. I think that having a dedicated secure OS is great and very usefull.

But is this a indication on how the developement model of OpenBSD is holding themselves back to much?

Decreased performance = increased costs for using something like that in a professional enviroment. There are plenty of companies using Linux and FreeBSD for important things and then they in return dedicate resources to the support of these OSes. Programmers and developers are going to seek a wider audiance, while openBSD isn't going to be that attractive if they have to make a choice in which platform to dedicate their time. Admins will eventuall drift towards using and playing around with OSes they use at work.

So that means that OpenBSD is going to have a hard time attracting "new blood" for more developement workforce and thus it goes in a downward spiral.

BTW does any companies or orginizations use openbsd for web serving anyways?

 

[Barnaby W. Füi]

I don't really have any beef with the benchmarks at all. They seemed mostly fair to me. OpenBSD performed the worst. Shouldn't that be expected? Their focus is security and cryptogaphy, and doesn't that generally slow things down? I don't think anyone will be switching OSes because of this benchmark, but I do think it was a good thing because 1. It helped NetBSD people fix some things that obviously they thought should be fixed, 2. It hopefully woke up some OpenBSD people to do the same sort of thing.

But even if the performance of NetBSD and OpenBSD stayed exactly the same, I still don't see any big problems. NetBSD is focused on portability, and OpenBSD is focused on security. FreeBSD is the one that's focused on performance, so leave the big monster web serving tasks to it or linux. I don't think that anyone says "use (Net|Open)BSD for EVERYTHING!", other than zealots. People generally say, oh, for that particular task, maybe xyz would be a good choice. And I don't imagine that the majority of people using any of these OSes even need high network scalability. Basically I think that the benchmarks were blown out of proportion, particularly the negative side of them. Lots of good came out of them and I think they were a win for everyone.

edit: Oh but I do think it was a bit out of line for the guy to make suggestions that people uninstall OpenBSD. Probably the biggest gripe I had about the whole thing. Just give us the numbers man, don't get preachy 😉

 

[chsh1ca]

Originally posted by: BingBongWongFooey
I don't really have any beef with the benchmarks at all. They seemed mostly fair to me. OpenBSD performed the worst. Shouldn't that be expected? Their focus is security and cryptogaphy, and doesn't that generally slow things down?

That is utterly unrelated to the tests done. Security/cryptography isn't slowing down system calls like mmap(), fork() open(), close(), and so forth. It is simply that the OBSD developers have put Security first, and speed second. That may not justify the incredibly bad results it got, but it does help explain it some. I look at it this way. If the tests were indeed wrong, the kernel devs of the various BSDs wouldn't have actually decided to try and fix some of the things that his tests exposed. I would assume the kernel developers know what they are doing.

 

[Barnaby W. Füi]

Originally posted by: chsh1ca

Originally posted by: BingBongWongFooey
I don't really have any beef with the benchmarks at all. They seemed mostly fair to me. OpenBSD performed the worst. Shouldn't that be expected? Their focus is security and cryptogaphy, and doesn't that generally slow things down?

That is utterly unrelated to the tests done. Security/cryptography isn't slowing down system calls like mmap(), fork() open(), close(), and so forth. It is simply that the OBSD developers have put Security first, and speed second. That may not justify the incredibly bad results it got, but it does help explain it some. I look at it this way. If the tests were indeed wrong, the kernel devs of the various BSDs wouldn't have actually decided to try and fix some of the things that his tests exposed. I would assume the kernel developers know what they are doing.

You just said yourself that it is related. Whether through technical issues or manpower issues, the focus on one thing is going to cause other things to lag behind a little more. It's not like OpenBSD advertises itself as "everything NetBSD, FreeBSD, and Linux have, PLUS a bunch of security crap!"

 

[n0cmonkey]

I wouldn't be surprised if the results were close enough to the same if the tests were done right. I am not saying it was REALLY unfair to anyone, except the people that expected something mroe scientific. This is a grreat start. I don't think this paper was worth giving a speech about in front of admins, developers, and PHBs.

The Linux developers deserve credit. It appears 2.6 is a big jump from everything before it. I wonder how 2.2 would stack up. Or, to put it another way, how much better is 2.6 than 2.4 was to 2.2?

The FreeBSD developers deserve some praise. 5.1 seems to perform much better than I expected. When some kind of -stable is released, I'll consider trying it out again.

NetBSD deserves a hell of a lot more than praise. I started typing something nasty about them, but decided it would be wrong. 😛 They saw problem, released fixes, and didn't bitch.

I only saw bitching from the non-developers in the BSD camp. The regular developers either did not comment, or made comments like the email I linked earlier. The OpenBSD users that did bitch obviously did the wrong thing. It doesn't help out the project in any way.

I have been considering installing NetBSD -current on my sparcstation the past couple of days. But I'm not sure if it will be worth the electricity 😉

On to OpenBSD's performance problems... UBC is in the works. It has been for a while. I imagine that is a tough thing to get working, and up to the usual scrutiny of the developers. SMP is also being worked on, but that has nothing to do with these benchmarks. I think UBC would help out OpenBSD's performance increadibly, but it is a slow process. I've considered downloading the UBC branch and trying it out, but I wouldn't know where to start. Maybe if I get my laptop bck sometime.

The OpenBSD project isn't dying, so we don't need any of that /. "BSD is dying crap." Several companies use OpenBSD for various tasks. Some even use it for high end webservers. If you need URLs, I'll find them for you. 😉 Many companies use it for perimeter and security devices. Experience with OpenBSD is how I got a job a couple of yeats ago 😉

And most people will not say Net/Open/Free BSD or Linux or Mac OS X are good for EVERYTHING. I am sure many of us on this board use various OSes. Some of the OpenBSD developers have been known to run other OSes (talked to one about Mac OS X before I bought my iBook). I even use Linux on my database server. I'd consider using OpenBSD if it was SMP capable, but Linux serves me well.

 

[chsh1ca]

Originally posted by: BingBongWongFooey
You just said yourself that it is related. Whether through technical issues or manpower issues, the focus on one thing is going to cause other things to lag behind a little more.

No, I said that the security stuff in each OS is unrelated to the benchmark. I was offering an explanation of why it was unrelated. These are basic fundamental operations that should work on every OS. They have to be at least working before any of the stuff related to security can be completed, if not started. The focus on security practices aside, OBSD still should be able to be on par with FreeBSD and NetBSD.

It's not like OpenBSD advertises itself as "everything NetBSD, FreeBSD, and Linux have, PLUS a bunch of security crap!"

No, but it IS like OpenBSD claims that it is slightly slower (small sacrifice in speed), but more secure. This is evidently not the case, as it is significantly slower (at the moment). Most people tout OpenBSD's security as a good thing, but when it comes down to it for businesses, a Linux 2.4/2.6 box can be just as secure, and run much faster. For scalable web services, the faster the underlying OS is at these operations, the less hardware you'll need in a large cluster for load balancing. Now, this may not matter to someone who is using 1 or 2 webservers, but to a large hosting corp. this performance difference could equate to dozens of boxes.

Originally posted by: n0cmonkey
I wouldn't be surprised if the results were close enough to the same if the tests were done right. I am not saying it was REALLY unfair to anyone, except the people that expected something mroe scientific. This is a grreat start. I don't think this paper was worth giving a speech about in front of admins, developers, and PHBs.

Definitely not. One of the benchmark's shortcomings is that, to my knowledge anyway, OpenBSD doesn't claim to be a high-performance high-scalability server solution. I could be mistaken. OpenBSD is designed more around security, and that is an entirely good idea. The approach should be encouraged IMO, since not enough people do it. In a way, the benchmark is flawed in the premise behind its approach.

The Linux developers deserve credit. It appears 2.6 is a big jump from everything before it. I wonder how 2.2 would stack up. Or, to put it another way, how much better is 2.6 than 2.4 was to 2.2?

I also wonder. Maybe you should write him and ask him to check. 🙂

I only saw bitching from the non-developers in the BSD camp. The regular developers either did not comment, or made comments like the email I linked earlier. The OpenBSD users that did bitch obviously did the wrong thing. It doesn't help out the project in any way.

In a way, it's understandable. I mean, how often on these very forums do you see people freaking out over a benchmark that shows NVidia/ATI doing poorly, or Intel/AMD? Brand loyalty at its best. 🙂

The OpenBSD project isn't dying, so we don't need any of that /. "BSD is dying crap." Several companies use OpenBSD for various tasks. Some even use it for high end webservers. If you need URLs, I'll find them for you. Many companies use it for perimeter and security devices.

LOL, yeah, they are kind of notorious for that. I personally have a hard time believing any non-microsoft OS has ever proven it can die once it has a decent foothold. Look as Novell's Netware. It's still used in large corporations, and in some respects is to this day significantly better than the network services that MS provides.

And most people will not say Net/Open/Free BSD or Linux or Mac OS X are good for EVERYTHING.

Of all the OSes, Mac OS X and Linux come closest to true general use OSes, simply by virtue of better vendor support (thanks IBM!) and that sort of thing.

 

[n0cmonkey]

Originally posted by: chsh1ca

Originally posted by: BingBongWongFooey
You just said yourself that it is related. Whether through technical issues or manpower issues, the focus on one thing is going to cause other things to lag behind a little more.

No, I said that the security stuff in each OS is unrelated to the benchmark. I was offering an explanation of why it was unrelated. These are basic fundamental operations that should work on every OS. They have to be at least working before any of the stuff related to security can be completed, if not started. The focus on security practices aside, OBSD still should be able to be on par with FreeBSD and NetBSD.

But there are security issues involved with the basics here. Read the part about UBC. OpenBSD doesn't have it. I believe all of the others do. A couple of things holding back UBC on OpenBSD: security and developer time. The same can be said about SMP 😉 There are plenty of security problems that have to be worked out before the OpenBSD team will feel comfortable releasing either of these MAJOR enhancements. And unfortunately, it does not have the support of Linux. Recently, the only major support it did have was taken away by a shortsighted committee.

It's not like OpenBSD advertises itself as "everything NetBSD, FreeBSD, and Linux have, PLUS a bunch of security crap!"

No, but it IS like OpenBSD claims that it is slightly slower (small sacrifice in speed), but more secure. This is evidently not the case, as it is significantly slower (at the moment). Most people tout OpenBSD's security as a good thing, but when it comes down to it for businesses, a Linux 2.4/2.6 box can be just as secure, and run much faster. For scalable web services, the faster the underlying OS is at these operations, the less hardware you'll need in a large cluster for load balancing. Now, this may not matter to someone who is using 1 or 2 webservers, but to a large hosting corp. this performance difference could equate to dozens of boxes.

But people do use OpenBSD. Companies do use OpenBSD. Here is proof. if you need it. 😉 I didn't check to see if NASA was actually listed on there, but they released a whitepaper on one of their uses. There have been reports about DoD use. It is in more places than you know. Plus, 99.9% of linux users owe Theo de Raadt in one way or another (think OpenSSH 😉). Some of the technologies coming out of the OpenBSD camp are amazing. NetBSD has some great projects of their own. The FreeBSD project is leading the pack on UFS2, IIRC. Each project has something to contribute. They may not be the best at everything, but they are there, and sometimes looking at things the Linux community ignores or does not see.

Originally posted by: n0cmonkey
I wouldn't be surprised if the results were close enough to the same if the tests were done right. I am not saying it was REALLY unfair to anyone, except the people that expected something mroe scientific. This is a grreat start. I don't think this paper was worth giving a speech about in front of admins, developers, and PHBs.

Definitely not. One of the benchmark's shortcomings is that, to my knowledge anyway, OpenBSD doesn't claim to be a high-performance high-scalability server solution. I could be mistaken. OpenBSD is designed more around security, and that is an entirely good idea. The approach should be encouraged IMO, since not enough people do it. In a way, the benchmark is flawed in the premise behind its approach.

The Linux developers deserve credit. It appears 2.6 is a big jump from everything before it. I wonder how 2.2 would stack up. Or, to put it another way, how much better is 2.6 than 2.4 was to 2.2?

I also wonder. Maybe you should write him and ask him to check. 🙂

I'm one of the users that keeps his mouth shut more often than not, no matter what some of you may think ;P

I only saw bitching from the non-developers in the BSD camp. The regular developers either did not comment, or made comments like the email I linked earlier. The OpenBSD users that did bitch obviously did the wrong thing. It doesn't help out the project in any way.

In a way, it's understandable. I mean, how often on these very forums do you see people freaking out over a benchmark that shows NVidia/ATI doing poorly, or Intel/AMD? Brand loyalty at its best. 🙂

And no matter how much it is condemned by any group, every group has fallen prey at one time or another.

The OpenBSD project isn't dying, so we don't need any of that /. "BSD is dying crap." Several companies use OpenBSD for various tasks. Some even use it for high end webservers. If you need URLs, I'll find them for you. Many companies use it for perimeter and security devices.

LOL, yeah, they are kind of notorious for that. I personally have a hard time believing any non-microsoft OS has ever proven it can die once it has a decent foothold. Look as Novell's Netware. It's still used in large corporations, and in some respects is to this day significantly better than the network services that MS provides.

I saw comments in one of the posts here that reminded me of the stupid BSD is dying crap on /. Had to say something 😉

And most people will not say Net/Open/Free BSD or Linux or Mac OS X are good for EVERYTHING.

Of all the OSes, Mac OS X and Linux come closest to true general use OSes, simply by virtue of better vendor support (thanks IBM!) and that sort of thing.

I think Mac OS X comes the closest to the best desktop ever. Linux would be third. Sorry guys. 😉 This is my opinion though, so don't take it too hard. A BSD user's opinion ain't worth much these days.

One thing I think should be looked at though, is how well each BSD has done without major corporate backing.

 

[n0cmonkey]

Here's the BSD is dying post I mentioned 😛 Might as well reply 😉

Originally posted by: drag
Well...

I agree that actual webserver performance using Apache would be usefull, it may not practicle for 2 reasons.

1. It would be just to darn expensive. He would have to have a dedicated webserver and a bunch of high-powered cllient machines to do it. (as you pointed out.) And probably would of had to had a couple other guys helping him.

I bet half of the people that posted in this forum have most of the hardware for it 😉

2.(and the bigger deal) People would freak out on him 10x as much as now. The version of Apache he used, php scripts, perl scripts, how big the files were. What network card he used, what drivers he used, etc etc. Every little detail, every little thing that could give one OS a artificial advantage over the other one. His e-mail box would get hammered.

I think how he did it would probably the best one guy could do in a resonable time alloted.

But don't you think he would want to make it the best test he could period? This wasn't one person's attempt at benchmarking, this was a presentation.

Now on a side note.

I like OpenBSD and all. I think that having a dedicated secure OS is great and very usefull.

But is this a indication on how the developement model of OpenBSD is holding themselves back to much?

Look at some of the interresting technologies coming out of the OpenBSD camp these days. I don't think they are holding back. Evolution of revolution, according to Theo anyways.

Decreased performance = increased costs for using something like that in a professional enviroment. There are plenty of companies using Linux and FreeBSD for important things and then they in return dedicate resources to the support of these OSes. Programmers and developers are going to seek a wider audiance, while openBSD isn't going to be that attractive if they have to make a choice in which platform to dedicate their time. Admins will eventuall drift towards using and playing around with OSes they use at work.

So that means that OpenBSD is going to have a hard time attracting "new blood" for more developement workforce and thus it goes in a downward spiral.

BTW does any companies or orginizations use openbsd for web serving anyways?

I posted a link in my post above this one. Atleast one of the developers (IIRC) has stated he is an admin at a large hosting company using OpenBSD.

The evolution over revolution approach is slow. It gets tiresome. But it has worked pretty well so far. Ok, the performance isn't great. Hopefully these benchmarks help the OpenBSD developers put bandaids on the stubs that used to be fingers and put out some more code to improve the performance. I've got my fingers crossed for UBC personally 😛

 

[chsh1ca]

Originally posted by: n0cmonkey
But there are security issues involved with the basics here. Read the part about UBC. OpenBSD doesn't have it. I believe all of the others do. A couple of things holding back UBC on OpenBSD: security and developer time. The same can be said about SMP 😉 There are plenty of security problems that have to be worked out before the OpenBSD team will feel comfortable releasing either of these MAJOR enhancements. And unfortunately, it does not have the support of Linux. Recently, the only major support it did have was taken away by a shortsighted committee.

Security issues or not, the fact is it performs significantly worse than the other BSDs. Ignore linux, since it is unrelated. The other BSDs are much better.

No, but it IS like OpenBSD claims that it is slightly slower (small sacrifice in speed), but more secure. This is evidently not the case, as it is significantly slower (at the moment). Most people tout OpenBSD's security as a good thing, but when it comes down to it for businesses, a Linux 2.4/2.6 box can be just as secure, and run much faster. For scalable web services, the faster the underlying OS is at these operations, the less hardware you'll need in a large cluster for load balancing. Now, this may not matter to someone who is using 1 or 2 webservers, but to a large hosting corp. this performance difference could equate to dozens of boxes.

But people do use OpenBSD. Companies do use OpenBSD. Here is proof. if you need it. 😉 I didn't check to see if NASA was actually listed on there, but they released a whitepaper on one of their uses. There have been reports about DoD use. It is in more places than you know. Plus, 99.9% of linux users owe Theo de Raadt in one way or another (think OpenSSH 😉). Some of the technologies coming out of the OpenBSD camp are amazing. NetBSD has some great projects of their own. The FreeBSD project is leading the pack on UFS2, IIRC. Each project has something to contribute. They may not be the best at everything, but they are there, and sometimes looking at things the Linux community ignores or does not see.

Umm, nothing I said had anything to do with who is using it. Of course people are using it. This kind of benchmarking had never been done before, and a lot of people I know were surprised that OBSD was so much slower. It was just generally assumed to be slightly slower or on par with linux. You are taking my comments as if they mean something that they really don't. OpenBSD has its place, and is in use, there is no denying that. It is somewhat irrelevant to what I said, since my comment was more on how things may change once peope look at the results.

And no matter how much it is condemned by any group, every group has fallen prey at one time or another.

Indeed.

Of all the OSes, Mac OS X and Linux come closest to true general use OSes, simply by virtue of better vendor support (thanks IBM!) and that sort of thing.

I think Mac OS X comes the closest to the best desktop ever. Linux would be third. Sorry guys. 😉 This is my opinion though, so don't take it too hard. A BSD user's opinion ain't worth much these days.

Are you saying that Windows would be second, or BSD would be second? If the latter, I have to laugh and not take you seriously... 🙂

One thing I think should be looked at though, is how well each BSD has done without major corporate backing.

BSD has had backing for a lot longer than linux has. Universities and the DoD have supported BSDs since far before linux had any such support. IBM only signed on recently, and it would be hard to classify vendors like RedHat, Mandrake, and SuSe as 'major'.

The end result of this whole thing is that the OSes in question can only be improved. 2.6 seems to be a huge improvement for linux, but I didn't need a benchmark to tell me that. It feels a LOT snappier than 2.4 was, especially for multitasking, and definitely blows XP away.

 

[n0cmonkey]

Originally posted by: chsh1ca

Originally posted by: n0cmonkey
But there are security issues involved with the basics here. Read the part about UBC. OpenBSD doesn't have it. I believe all of the others do. A couple of things holding back UBC on OpenBSD: security and developer time. The same can be said about SMP 😉 There are plenty of security problems that have to be worked out before the OpenBSD team will feel comfortable releasing either of these MAJOR enhancements. And unfortunately, it does not have the support of Linux. Recently, the only major support it did have was taken away by a shortsighted committee.

Security issues or not, the fact is it performs significantly worse than the other BSDs. Ignore linux, since it is unrelated. The other BSDs are much better.

Better? Maybe. Faster in these benchmarks? Definitely.

No, but it IS like OpenBSD claims that it is slightly slower (small sacrifice in speed), but more secure. This is evidently not the case, as it is significantly slower (at the moment). Most people tout OpenBSD's security as a good thing, but when it comes down to it for businesses, a Linux 2.4/2.6 box can be just as secure, and run much faster. For scalable web services, the faster the underlying OS is at these operations, the less hardware you'll need in a large cluster for load balancing. Now, this may not matter to someone who is using 1 or 2 webservers, but to a large hosting corp. this performance difference could equate to dozens of boxes.

But people do use OpenBSD. Companies do use OpenBSD. Here is proof. if you need it. 😉 I didn't check to see if NASA was actually listed on there, but they released a whitepaper on one of their uses. There have been reports about DoD use. It is in more places than you know. Plus, 99.9% of linux users owe Theo de Raadt in one way or another (think OpenSSH 😉). Some of the technologies coming out of the OpenBSD camp are amazing. NetBSD has some great projects of their own. The FreeBSD project is leading the pack on UFS2, IIRC. Each project has something to contribute. They may not be the best at everything, but they are there, and sometimes looking at things the Linux community ignores or does not see.

Umm, nothing I said had anything to do with who is using it. Of course people are using it. This kind of benchmarking had never been done before, and a lot of people I know were surprised that OBSD was so much slower. It was just generally assumed to be slightly slower or on par with linux. You are taking my comments as if they mean something that they really don't. OpenBSD has its place, and is in use, there is no denying that. It is somewhat irrelevant to what I said, since my comment was more on how things may change once peope look at the results.

Do you think these benchmarks are going to change that many minds?

And no matter how much it is condemned by any group, every group has fallen prey at one time or another.

Indeed.

Of all the OSes, Mac OS X and Linux come closest to true general use OSes, simply by virtue of better vendor support (thanks IBM!) and that sort of thing.

I think Mac OS X comes the closest to the best desktop ever. Linux would be third. Sorry guys. 😉 This is my opinion though, so don't take it too hard. A BSD user's opinion ain't worth much these days.

Are you saying that Windows would be second, or BSD would be second? If the latter, I have to laugh and not take you seriously... 🙂

Upon reflection, I take back this statement. I was thinking Windows at the time, but I have beer in me now. BSD, in my opinion, is the best. But, if BSD did not exist Linux would probably be my choice.

One thing I think should be looked at though, is how well each BSD has done without major corporate backing.

BSD has had backing for a lot longer than linux has. Universities and the DoD have supported BSDs since far before linux had any such support. IBM only signed on recently, and it would be hard to classify vendors like RedHat, Mandrake, and SuSe as 'major'.

BSD had support before Linux was concieved. Then AT&T got involved. The support stopped for quite a while.

The end result of this whole thing is that the OSes in question can only be improved. 2.6 seems to be a huge improvement for linux, but I didn't need a benchmark to tell me that. It feels a LOT snappier than 2.4 was, especially for multitasking, and definitely blows XP away.

Agreed. We all have good things to look forward to, whether you are into Linux, BSD, or just *nix in general. NetBSD and OpenBSD have both shown improvement. And I don't know about anyone else, but one good thing about these benchmarks is that I am seriously considering trying the other BSDs more than ever.

 

[drag]

The OpenBSD project isn't dying, so we don't need any of that /. "BSD is dying crap." Several companies use OpenBSD for various tasks. Some even use it for high end webservers. If you need URLs, I'll find them for you. Many companies use it for perimeter and security devices. Experience with OpenBSD is how I got a job a couple of yeats ago

Good to here that. We realy need stuff like OpenBSD. I realy don't care how fast it is compared to Linux of FreeBSD, as long as it is fast enough. I probably wouldn't use it for a webserver, but for something that I need to set up in a hostile enviroment and need as little administration overhead as possible, OpenBSD rocks. (like a security station, or a firewall, simple router, or a terminal server or something..)

BTW does OpenBSD support routed (edit: I ment ROUTING) protocols and stuff like that, for large network convergance and event horizon crap like that? How about border gateway protocols?

 

[n0cmonkey]

Originally posted by: drag

The OpenBSD project isn't dying, so we don't need any of that /. "BSD is dying crap." Several companies use OpenBSD for various tasks. Some even use it for high end webservers. If you need URLs, I'll find them for you. Many companies use it for perimeter and security devices. Experience with OpenBSD is how I got a job a couple of yeats ago

Good to here that. We realy need stuff like OpenBSD. I realy don't care how fast it is compared to Linux of FreeBSD, as long as it is fast enough. I probably wouldn't use it for a webserver, but for something that I need to set up in a hostile enviroment and need as little administration overhead as possible, OpenBSD rocks. (like a security station, or a firewall, simple router, or a terminal server or something..)

BTW does OpenBSD support routed protocols and stuff like that, for large network convergance and event horizon crap like that? How about border gateway protocols?

Don't know off hand. There may be a project or two out there providing that support. There was some VRRP support through third party sources, but I don't know if that project is still active since CARP came about.

 

[n0cmonkey]

Probably more good news for Linux (and Sun).

 

[drag]

BTW, I looked it up and appariently OpenBSD is able to use Zebra, which is a set of GPL software to replicate the functioning of cisco routers, I guess. However it's been dead, no updates for close to a year, so a fork, Quagga software has been started to continue the project.

 

[chsh1ca]

Originally posted by: n0cmonkey
Do you think these benchmarks are going to change that many minds?

Judging by the reactions of both the NetBSD and OpenBSD devs, it already has. In terms of userbase, perhaps not, since a lot of BSD users tend to be loyal to the software, whether right or wrong.

Upon reflection, I take back this statement. I was thinking Windows at the time, but I have beer in me now. BSD, in my opinion, is the best. But, if BSD did not exist Linux would probably be my choice.

LMAO, no offense, but I find that funny.

BSD had support before Linux was concieved. Then AT&T got involved. The support stopped for quite a while.

The point being that Linux has outgrown BSD in terms of use with little corporate support, and has been growing in the area of corp. support quite steadily for some time now, and with two bigger players in the industry in the last two years picking up support for it, and deciding to promote it, I'd say that's quite the accomplishment for something that was conceived as a replacement for minix, and isn't in any way genetically tied to other code. Whether that is more of an accomplishment than redevelopment, I'm not entirely sure. Not that it particularly matters.

Agreed. We all have good things to look forward to, whether you are into Linux, BSD, or just *nix in general. NetBSD and OpenBSD have both shown improvement. And I don't know about anyone else, but one good thing about these benchmarks is that I am seriously considering trying the other BSDs more than ever.

I may have to give the BSD's another shot on my spare box. 🙂

 

[n0cmonkey]

Originally posted by: chsh1ca

Originally posted by: n0cmonkey
Do you think these benchmarks are going to change that many minds?

Judging by the reactions of both the NetBSD and OpenBSD devs, it already has. In terms of userbase, perhaps not, since a lot of BSD users tend to be loyal to the software, whether right or wrong.

I am not sure it changed the minds of developers, just gave them something specific to look at. I don't think there is a right or wrong here. It's all about what works for me. Or for you I guess. 😉

Upon reflection, I take back this statement. I was thinking Windows at the time, but I have beer in me now. BSD, in my opinion, is the best. But, if BSD did not exist Linux would probably be my choice.

LMAO, no offense, but I find that funny.

None taken. That was partially what I went for. Usually the BAC isn't so low, but these are tough times 😉

BSD had support before Linux was concieved. Then AT&T got involved. The support stopped for quite a while.

The point being that Linux has outgrown BSD in terms of use with little corporate support, and has been growing in the area of corp. support quite steadily for some time now, and with two bigger players in the industry in the last two years picking up support for it, and deciding to promote it, I'd say that's quite the accomplishment for something that was conceived as a replacement for minix, and isn't in any way genetically tied to other code. Whether that is more of an accomplishment than redevelopment, I'm not entirely sure. Not that it particularly matters.

Linux is very tied to other code. The GNU base is quite important. But most free OSes are kind of requiring that right now.

Agreed. We all have good things to look forward to, whether you are into Linux, BSD, or just *nix in general. NetBSD and OpenBSD have both shown improvement. And I don't know about anyone else, but one good thing about these benchmarks is that I am seriously considering trying the other BSDs more than ever.

I may have to give the BSD's another shot on my spare box. 🙂

If I can use Linux and not bitch too much about it, just about anyone can try a BSD 😉