BrandZ, which, quoting from the OpenSolaris BrandZ Community webpage, is a framework that extends the Solaris Zones infrastructure to create Branded Zones, which are zones that contain non-native operating environments, integrated into the OpenSolaris mainline OpenSolaris mainline kernel last night.
running Brandz. The rest I can compile and run at native speeds.
I think you missed the point.
BrandZ is not an emulator. The user space code runs natively. The lx branded zone implements the linux syscall layer in parallel to the Solaris one.
With regard to acrobat, we do already have evince.
BrandZ will also allow us to DTrace linux applictions with no change whatsoever to the application.
alan.
That’s awesome.
Big thanks to the engineers who made it work.
Correct me if I’m wrong but isnt BrandZ simular to FreeBSD’s Linux layer? A layer the implements system calls found in the Linux kernel and translates them to call understood by the native kernel?
Is that how the Linux apps are able to run? Does BrandZ implement all the system calls found in the latest 2.6.x kernel or is it like FreeBSD’s which only supports 2.4.x system calls?
If the latter, it will mean Solaris may suffer the same limitations as FreeBSD when it comes to running branded Linux apps.
Exa: Transgaming’s Cedega doesn’t work on FreeBSD with the Linux layer.
That’s exactly what I thought when I read this. The lx brand they discuss would be quite similar to the FreeBSD Linux binary compatibility. The key difference here though is that everything runs inside a protected environment. To do this in FreeBSD, you would have to install the linux-base ports/packages inside of a FreeBSD jail and run your Linux applications inside of the jail. Only then would you receive the similar effect of security, isolation, and binary compatibility.
Another perhaps not-so important point to note is that the BrandZ system is designed to have several different brands developed for it. I say not-so important because I really see little reason to create other brands aside from Linux. A FreeBSD brand perhaps, but still I see that as a very niche market (not FreeBSD and it’s users, but those that would use FreeBSD in a BrandZ environment).
Now what I could see as having quite a demand would be a MacOS brand. Though this can only be viable when the BrandZ system supports 64-bit binaries as all the new Intel Macs are 64-bit.
The initial brand is for Linux 2.4 applications but the concept can be extended to support 2.6 and future versions as well. The important thing is the infrastructure has been made available in OpenSolaris so that new brands can be developed.
As one of the other posters said, what’s particularly nice about this sort of virtualization is the besides the system call interposition that takes place, since it runs within a container/zone, the complete Linux userland “boots” up just like a regular system. This differs from other technologies which are more focused on running a specific application.
Finally, the use of DTrace to debug and tune Linux applications should be very useful.
An isolated environment would add extra security focus as well. Pretty neat. Does this mean users would not need to brand individual applications like FreeBSD?
I heard Nexenta will support BrandZ once its implemented in ON; it seems that will happen in build 49. Nexenta is the most impressive OpenSolaris distribution but is too buggy (at the moment) for general use.
I’m not familar with the FreeBSD solution but there’s no need to brand individual applications. In general, almost all well-behaved and some ill-behaved applications should work as-is. The ones that may have issues are ones that rely on a particular Linux device driver or some of the more esoteric features. If you want to get more details, please read the BrandZ design document which was posted on the BrandZ community page several months ago
http://www.opensolaris.org/os/community/brandz/design/
I really believe that OpenSolaris has all the ingredients of success:
* innovation – ZFS, DTRACE, Zones/Contains/Brands, resource control (yes, cpu caps are coming)
* open – open source, open standards generally, can be made to play well with industry favorites, from x86 to grub, from ndivia opengl to realplayer!
* quality development process – the design and implement process is a very much more considered and thoughtful than the linux process. the code quality is much higher. however, unlike the free BSDs, opensolaris has momentum and real results.
* performance – eg new network stack (project fireengine) – aim to saturate a 1GB line on server hardware.
in the interests of rigour and debate – i’d welcome reasons as to why anyone wouldn’t use OpenSolaris for their next project …
update: one negative aspect to opensolaris and the official solaris derived from it is the userland. sure they have very well tested and standards compliant alternatives … /usr/bin/ps vs /usr/ucb/ps and so on … but try to compile anything out of the box and its a pain. only recently have they started including useful tools in the /usr/sfw hierarchy including ssl and ldap libraries. still no readline libraries though.
Edited 2006-09-13 12:58
The main reason I have for not trying it out is that I can’t work out how to download it without registering with Sun. So much for being “open”, eh?
Also, for the record, an unqualified statement like “the code quality is much higher [than in Linux]” is bound to wind people up…
And how do you determine that because you have to register that Solaris/OpenSolaris is not open? As I have pointed out before (along with a number of Sun employees) registering does not put you on an spam list or some other “evil” thing?
All of the e-mail I get from Sun is voluntary, I wished I could say the same thing for Microsoft.
Edited 2006-09-13 16:42
I don’t mind giving my name but Sun needs to get off that hokey number of PCs for Development, number of PCs for Commercial use, number of PCs for farting around. I always enter 100 for each of the categories…..that’s how Sun can claim 100 million downloads of Open Solaris.
I don’t mind giving my name but Sun needs to get off that hokey number of PCs for Development, number of PCs for Commercial use, number of PCs for farting around. I always enter 100 for each of the categories…..that’s how Sun can claim 100 million downloads of Open Solaris.
No, they don’t need to “get off of it.” The only thing screwing up the numbers are those that are intentionally dishonest or don’t enter the proper numbers. It’s perfectly within their right to ask you to take a few minutes of your time to tell them how many computers you want to use the software on. After all, they are giving it away for free!
It’s perfectly within their right to ask you to take a few minutes of your time to tell them how many computers you want to use the software on. After all, they are giving it away for free!
It’s annoying and that’s a sure fire way to dissuade users to something you’re giving away for free. Why should sun care anymore?. This kind of crap is as effective as that Linux Counter project. Asking for meaningless statistics is a waste of my time. Make it optional and perhaps reward someone for filling in that part of the questionnaire by giving them points towards Sun support or purchasing Sun hardware.
Give me a break! You can’t take the time to answer a few questions but you can take the time to download and use the OS. I don’t see what the problem is, just as binarycrusader said “After all, they are giving it away for free!”.
And you actually expect Sun to give you consideration when you won’t give them any? Get real! The only way I could possibly see Sun giving people breaks like you propose is if you were registering thousands of machines.
And how do you determine that because you have to register that Solaris/OpenSolaris is not open?
Do I have to register to download Debian? Do I have to register to download FreeBSD?
If you don’t want to register, you could download an OpenSolaris distribution from another vendor than Sun (for example http://gnusolaris.org, http://schillix.berlios.de/, …)
If you don’t want to register, you could download an OpenSolaris distribution from another vendor than Sun (for example http://gnusolaris.org, http://schillix.berlios.de/, …)
Nexenta seems to be Ubuntu with the Solaris kernel. That’s an interesting project, but it’s not the real deal. Schillix, from what I can tell, doesn’t even come with a GUI.
Why isn’t there just a mirror with the full OpenSolaris DVD ISO on, available for anyone to download without restriction, as with the BSDs and all the myriad versions of Linux?
What makes you think GNU/Solaris (Nexenta) is any less ‘the real deal’ than what you seem to refer to as ‘regular Open Solaris’ eventhough it’s not specifically mentioned?
Is it because it’s Open Solaris plus a lot of goodies from Ubuntu, e.g. apt-get package management and the look and feel of ‘Ubuntu Gnome’ 2.14.x?
It’s got DTRACE, Zones, ZFS …
I admit it might be nice to have a true Open Solaris rig with X.org 7.1 and Gnome 2.16 but for now, Nexenta is pretty close…
Why isn’t there just a mirror with the full OpenSolaris DVD ISO on, available for anyone to download without restriction, as with the BSDs and all the myriad versions of Linux?
Because Solaris isn’t “just like” all the other Operating Systems. SUN has purposefully avoided making an official OpenSolaris distribution because the community doesn’t want SUN to be the one to provide it, not only that, it’s part of the goal of the OpenSolaris project that SUN does *not* provide it. Without this as a guideline, you might as well download Solaris and be done with it!
The main reason I have for not trying it out is that I can’t work out how to download it without registering with Sun. So much for being “open”, eh?
You can download the source without registering.
Solaris Express is *NOT* OpenSolaris, it is based on OpenSolaris technology, however. If you want an OpenSolaris distribution, try Nexenta, SchilliX, etc:
http://www.gnusolaris.org
http://www.schillix.de
OpenSolaris is perfectly open.
Also, for the record, an unqualified statement like “the code quality is much higher [than in Linux]” is bound to wind people up…
So if you do not agree, you are telling us that “the code quality is much higher on linux”.
Why your opinion should be more qualified than theirs ?
Just because you like linux and not solaris ?
…and here we go again with the “omg Solaris rocks, everything else sux!!1” posts.
And you’re not? Where has any poster for this particular piece said anything along the lines of “omg Solaris rocks, everything else sux!!1” The comment by project_2501 might be considered by some as “rah rah”, but he also has some concerns that he voices about the userland.
Or as usual you are puked out that Sun and the OpenSolaris project is making good on their promises, or that there is another article on OSAlert about the OS you hate.
Or as usual you are puked out that Sun and the OpenSolaris project is making good on their promises, or that there is another article on OSAlert about the OS you hate.
Oh, please. OS I hate? Think in more than one dimension.
but he also has some concerns that he voices about the userland.
Not just Rah Rah but also says point blank that nothing else is touching Solaris for code quality, etc. etc. which is utter tripe.
Well, show us three things you like about Solaris, and I do think in more than one dimension. I don’t think you do, but that is my opinion.
And while you are at it, can you backup your statement about the code quality of Solaris is tripe.
Hey it could be worse. It could be yet another article about some irrelevant software company that hasn’t created an OS in 6+ years, yet gets 6 articles on the front page about an OS that doesn’t exist yet.
Well, show us three things you like about Solaris, and I do think in more than one dimension. I don’t think you do, but that is my opinion.
Unique things? ZFS, code compatibility with older releases, excellent online documentation.
And while you are at it, can you backup your statement about the code quality of Solaris is tripe.
Easy. A fantastically bad bug in the network stack of 2.6 took Sun about a year to address. The kernel NFS server would bind in a quasi-random fashion to whatever IP address it wanted to (big issue with multiple ether ports and/or multiple IPs on a single port), making host-based NFS security (such as it is) more or less impossible because the server would answer over the ethernet using whatever IP address it had picked – if it was the wrong one, too bad for you.
A startlingly bad bug which showed a very myopic approach to network stack programming (to support their 4-port ganged ether card, as I remember) from the kernel devs and which took months for them to even acknowledge as a show-stopper. It then took Sun about a year to fix, which they did only for customers with a contract until 2.7 came out, IIRC.
So please, no lectures on Sun’s code quality being higher than anyone else’s. Some of us know better.
Now how about pointing out problems with code in the current release, I (and other readers) could care less about what happened six to eight years ago.
Now how about pointing out problems with code in the current release, I (and other readers) could care less about what happened six to eight years ago.
Hah… caught in the lie, you simply divert to some other argument. I get it, but it’s not working on me.
Actually, Robert wasn’t trying to divert the topic. You made a claim about Solaris 2.6. That’s a release which Sun end-of-lifed quite some time ago. This particular article is about OpenSolaris, ie, the *current* codebase. If you have a problem with demonstrating a code quality issue in the current release (which is freely available) then don’t claim that Sun’s code is lacking quality.
Put up, or shut up. And make sure your example is relevant to the topic at hand.
Shaman, do you happen to have the Sun BugID for the issue in question? I’m curious and would like to look it up.
Certainly some things fall through the cracks at times and it does take often too long to get issues resolved. That said, I believe that Sun is very good at delivering a quality operating system – we put an emphasis on regression testing. It doesn’t catch everything – obviously – but a large share of the issues are resolved early on which is what I think the original poster was implying about the code quality.
By the way, there have been a lot of work over the years in source address selection. It’s by no means perfect but it’s a lot better than it used to be.
I guess we don’t want to hear about Sun’s failures, do we, astroturfers? If you were the real deal, you’d have posted something about Sun learning from its mistakes and getting better, with examples. They have, you know… before Solaris 10, Sun’s OS was like working with a museum UNIX.
I understand your frustrations, though. It’s easier, as astroturfers who don’t actually work with Solaris first-hand, to simply say “oh that was then, this is now, put up or shut up!” If only I’d hold my peace, you could do your jobs and hype up Sun mostly uncontested, here.