Wednesday, February 27, 2008

FreeBSD 7.0

There is an interview with several FreeBSD developers on what's new in version 7.0

I have mixed feeling about FreeBSD - you look thru what's new and you got a feeling it's just catching up with other OS'es with no innovation on its own. IPMI support? 1:1 threading model? Journaling for a file system? - and not even "real" one... Maybe they did managed to get slightly better performance for a specific workloads but the question is - why would I need FreeBSD?
Then you have little ISV support - try to get Oracle libraries for FreeBSD.

In x86 market you even have already a big market share or you innovate to get one, you add some new value - frankly I can't find any added value in FreeBSD comparing to Solaris or Linux on x86. Sure, one can find some small niche where it makes sense, but in general one will be much more happy with Solaris or Linux.

Sadly (or maybe not?) FreeBSD is becoming more and more an OS for hobbyst - or maybe it has always been....

10 comments:

Anonymous said...

Sadly, the interview was not written as an article; it's mostly developer comments concatinated together, without resulting in a clearer picture.

I recommend you have a look at the slides of Kris Kennaway's FOSDEM talk: 7.0 and beyond - this provides a much better overview at what has been achieved, as well as what the current efforts are.

Anonymous said...

I think you're being a tad unfair. FreeBSD has licensing advantages over Linux (and arguably a more considered development process), and it will fit onto a very small device - unlike Solaris. I have it running on a Vortex-based system from a small CF drive. What's the chance I can get an OpenSolaris system with a (cut down, admittedly) JRE onto a 128MB CF, and running happily in a 128MB RAM system with readonly root? It also seems to outperform Solaris, at least on small systems.

Noah said...

I honestly prefer the way *BSDs are handled as far as submission of new code and features. As was also mentioned, Licensing considerations are another boon.

I'm more of an OpenBSD guy myself for both server and desktop use. I ditched Linux for FreeBSD back in '99 and occasionally look back, but for the most part, I've used Open since the turn of the century as I didn't like where FreeBSD 5 was headed.

I've recently given FreeBSD 6.3 a shot on the desktop (yes, as a hobbiest) and it's come a long way since 5.1.

For an enterprise UNIX server, I will usually look to Solaris, OpenBSD or AIX before Linux, though. And Windows? We won't talk about that.

milek said...

I know there's more that just in that interview - still nothing compeling for a desktop or server why anyone should go for FreeBSD instead of Windows, OSX, Linux or Solaris.

I like FreeBSD license more than Linux's GPL - agree. But the thruth is for majority of people as long as it is free and open source they don't care at all. NetApp and other companies which are doing propertiary platforms and want to keep it close sorce so they can charge extra - they care. But users? Most of them not.

Now, I agree that FreeBSD has its niche, one of them is problably embeded marktet. When it comes to servers or desktop it just doesn't make sense.

Performance - I've been doing some quick BIND performance tests on FreeBSD, Linux, Solaris on x86 multi-core system - FreeBSD is by large margin worst one - ok, it was 6.3 - maybe in the latest SMP is addressed better.

But still, on small x86 servers (4-8 cores) usually differences in performance between different operating systems in real world aren't big enough to be concern with. What matters more is added value - easy of management, applications availability, etc.

If you buy large disk array with a support - do you think FreeBSD will be supported? If not - why bother? You will get much better storage management with Solaris anyway, and it will be supported.

If you need an enterprise clustering - these days Sun Cluster is for free, both x86 and SPARC.

Your applications needs Oracle - you're in trouble on FreeBSD. And so on...

Then FreeBSD was first in a market with Jails, Sun "copied" technology and named it Zones - compare them, it's much more mature and feature rich on Solaris than on FreeBSD.

Anonymous said...

Sorry Milek, but you must be living in a dream world

FreeBSD *is* an hobbyist OS, and a very good one IMHO

to be honest, I'm pround of all the *BSD's, which are all very fine pieces of software

keep in mind that there's no multi-million dollar company behind the *BSD's like Sun (Solaris), or several of them like RedHat, Novel, IBM, etc. (Linux)

also keep in mind that most people working on *BSD do this in their spare time, and those people managed to keep up with Solaris and Linux

I have no idea why you would need FreeBSD, but I think the world wouldn't be a better place without it

Trochej said...

I have seen a post about running OpenSolaris in that constrains. There is a new OpenSoalris distribution based on Damn Small Linux, with ISO of 90 MB. I believe, it can be made even smaller.

Anonymous said...

I also hope that Solaris will be Damn Small one day ;-)

I mean I was shocked by this "just install everything" mentality of the solaris community

as a long time *BSD user it's just a big NO-NO to have things like X11 (+ GNOME *shrug* and StarOffice) on a server...

also think of the security, the less software is installed on your server, the less you'll have to patch sooner or later

well that could be a reason why FreeBSD is leading the uptime records on netcraft, and not solaris/linux...

milek said...

I've nevr said it's not good - it is a good system. I believe in open competition so I agree - it's good we have options.

I was just making a point, from a sys admin perspective working in a company - on a server there is no point in using FreeBSD comparing to Solaris or Linux due to reasons I stated.

Re Solaris crowd mentality - I don't know, but systems I do install via jumpstart are trimmed down - no X11, GNOME, etc...

Anonymous said...

Indeed there is a new small Solaris distro. But I got a tiny FreeBSD installation using the standard installer 6.2 and just turning on readonly root.

Admitedly, I hacked the bloated JRE in a way that's probably not really allowed by the current license - I unzipped the standard library and rm'd all the GUI, corba, remoting, and multimedia classes and then jar'ed is back up, and I rm'd the corresponding native libs.

Its a lot easier than futzing with aufs and assorted Linux build mechanisms.

Probably could get it smaller still if was prepared to try a minimal mono. I'd use Lua but I like the facilities in Java and C#.

I will certainly try that USB Solaris though.

I don't mean to imply that I'd use FreeBSD as a production server over Solaris, or indeed over Red Hat or SUSE. Mostly the issue there is indeed what you can get ISV support for - its not a technical issue at all. I do hope that FreeBSD keeps getting better though, if only to keep Linux honest, as it were.

c said...

it is simply a matter of what your needs are. if you prioritize new features and gizmos over stability, and you do not have the time or inclination to learn how to configure and tune your boxes, then FreeBSD may not be the best choice for you. however this in of itself does not suggest that FreeBSD is a hobbyist OS.
in fact, the definition of a hobbyist OS is one that prioritizes new features and gizmos over stability, which is the exact opposite of FreeBSD.
i choose FreeBSD specifically because i work in a production environment and i need an OS that does exactly what it says it can do; no more, no less. i simply cannot afford a machine that is anything but rock solid.