Sun Microsystems will create an open-source project around its Solaris 10 operating system by the end of the year, company executives said Monday. Also, Jonathan Schwartz showed 32-Way UltraSPARC chip that runs Solaris 10.
Solaris 10 to be Open Source
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Nice. I hope the licensing allows porting stuff over to Linux.
Nice. I hope the licensing allows porting stuff over to Linux.
Why not just use Solaris?
Why not just use Solaris?
Becuase Linux is free. And because GNU tools are much better than Solaris, or and other UNIX (with the exception of AIX maybe). And because all Linux is missing to put UNIX to rest is some of UNIX’s scalability.
UNIX scalability + GNU tools + 2.6 == The shizzle
it´s yet to be seen whether the licence will be OSI approved… Will it be able a Debian Solaris port?
> Why not just use Solaris?
Because I like the idea of Linux succeeding all the different proprietary Unix versions of old, personally.
Becuase Linux is free. And because GNU tools are much better than Solaris, or and other UNIX (with the exception of AIX maybe). And because all Linux is missing to put UNIX to rest is some of UNIX’s scalability.
UNIX scalability + GNU tools + 2.6 == The shizzle
Personally, I prefer the BSD tools :-p
Anyway, I’m looking forward to Solaris being open source…hopefully Linux and Solaris can both grow and “learn” from each other.
I doubt Solaris will be compatible with the GPL. They probably will have some kind of license term forcing any code written by others for Solaris to remain open but allowing Sun the capability of “owning” that code and closing the source as they see fit. Thus if any code gets transferred to the Linux kernel, in essence that person will have to give the kernel ownership to Sun and allow them to close the source as they wish. Of course, the Linux license is not compatible with this so code will not be going from Solaris to Linux or vice versa.
There is nothing that Solaris “could learn” from Linux. The only things that Linux has that Solaris is lacking are a buttload of drivers, many of which are of questionable quality at the moment.
Solaris is far more tested, scalable and secure, and will remain so for the next few years at any rate. In reality, the only thing that Linux has going for it is a free, open source license. It remains to be seen just what kind of license Solaris 10 will have, but if the unthinkable happens, and it’s released under (for example) the BSDL, then the drivers are going to be the only thing that Linux will have over Solaris, and you’d better believe that the Solaris folks will be doing their damndest to port over drivers from the various BSDs, as well as forking out the cash to just get some of the bigger hardware vendors to make proper drivers.
I’m not however, holding my breath here, hoping that they will release it under one of the less restrictive open source licenses…
Becuase Linux is free.
And so is anything released under an OSI approved license, which Sun has already said Solaris will be.
And because GNU tools are much better than Solaris, or and other UNIX (with the exception of AIX maybe).
Like others on this thread, I prefer the BSD userspace for a variety of reasons which would require a tool-by-tool comparison, which I won’t go into here.
That said, it’s doubtful if we’d see the merger of GPL code into the Solaris userspace, whereas due to the freedom afforded by the BSD license it’s entirely possible we’d see BSD code merged in.
And even so, Sun has made significant strides with their userspace in Solaris 9 & 10. If you are judging from Solaris <= 8, or even from Solaris 9, you should really check out Solaris 10, because your judgements really aren’t applicable anymore.
And because all Linux is missing to put UNIX to rest is some of UNIX’s scalability.
Scalability, resource partitioning/configurability, profiling support, advanced security (RBAC and ilk), virtualization support, and high availability features. Linux can have each of these through it’s bazaar approach with 3rd party add ons, but they certainly aren’t integrated into one cohesive package like they are with Solaris (although distributions like Fedora are picking up many of these such as SELinux).
The “it just works” philosophy means that everything is there and ready to go when you need it and you don’t need to do a lot of work to get things going, especially on a system-by-system basis. For me, the amount of work required to get certain things going in Linux when I’m maintaining several dozen systems (or hundreds if you count cluster nodes) is death by a thousand pinpricks.
And because GNU tools are much better than Solaris
Except for Solaris’ p-tools, which are miles beyond their Linux equivalents. Not to mention trapstat, lockstat, kstat, nfstat, busstat, cpustat, etc, or the mother of them all: DTrace. And many Solaris commands (ls, du, df) now support the -h flag, so I’m in heaven.
Solaris is far more tested, scalable and secure, and will remain so for the next few years at any rate.
Agreed. The only major change that has been made in Solaris in recent history has been the new 1:1 threading implementation in Solaris 9, which I understand was done because the M:N implementation was simply messy and unmaintainable.
Meanwhile Linux has thrown the baby out with the bathwater and rewritten many major kernel subsystems with each version release, most notably the VMM. This has been terribly problematic for me on certain systems, as I’m experiencing VM stability issues in 2.6 kernels in my higher load servers which remind me of the same problems I was experiencing in early 2.4 series kernels. I have since downgraded those systems back to 2.4 until 2.6 stabilizes, but from my experience 2.6 has been growing progressively less stable with each release.
Meanwhile my Solaris/SPARC servers are 100% aproblematic. While most of them aren’t under the same degree of load that my Linux servers are under, it’s really nice that they don’t need babysitting.
Because I like the idea of Linux succeeding all the different proprietary Unix versions of old, personally.
And this is a good thing? Yuck. Give me Solaris anyday.
UNIX scalability + GNU tools + 2.6 == The shizzle
You know there is nothing stopping you from using the gnu tools on Solaris. Though Sun’s compiler produces smaller more efficient binaries on SPARC than gcc. But if you must have the gpl ad nauseum or just don’t have the cash (which I can understand) gcc does an ok job.
I doubt Solaris will be compatible with the GPL
Possibly not, though I wouldn’t be too sure it won’t be.
The only things that Linux has that Solaris is lacking are a buttload of drivers, many of which are of questionable quality at the moment.
You are completely right there. I spent hours trying to get Linux working on an old armada laptop. The ide controller sort of worked but any time I’d try hdparm the machine would freeze up. Sound worked but I had to mute line in, the mic, and also line out. Even then I heard noise. I still have nightmares with that stupid sound clip of Torvalds pronouncing Linux. In freebsd, sound worked perfectly, but the ide controller wasn’t supported so I had no dma. I ended up selling the laptop so it’s kind of a mute point though.
If solaris go the restricted license way no one will bother with it ? if they go a good Opensource License yes they could use open source and port it to Solaris, but then so could the linux crowd
and im sure we all know who will do it faster, and maybe better ?
but that said solaris maybe licensed in chunks, say the kernel with more restricted license and the rest more open ?
Personally im sure Linux will over come any stuff tries to throw at it get it off track of domination. Even the BSD (FeeBSD) is making in roads in alot of companies, im not saying that linux is the reason for this, but Red Hat are making in roads, and once companies get a taste of opensource they are more willing to try other OS e.g BSD, so Solaris is got a uphill struggle imo
Snake
I am seeing quite a few misinformed comments which claim that linux doesnt scale well.
Please take a look at Altix where linux runs happily on 512 CPUs without any trouble.
Linux is also going to run on the NASA 10000 CPU Project Columbia machine.
I am not sure why people would still want to keep disparaging linux wrongly like this.
“That said, it’s doubtful if we’d see the merger of GPL code into the Solaris userspace, whereas due to the freedom afforded by the BSD license it’s entirely possible we’d see BSD code merged in.”
Thats just a way of saying that it’s easier to put BSD code into solaris no matter what license they use.
There are some things that Linux does very well…even better than Solaris. As an example, the performance of Solaris’ UFS has traditionally been considered to be very poor. A port of ext3, XFS, ReiserFS, or JFS would certainly be a nice additional option.
*cough* VxFS *cough*
There’s also cluster filesystems to consider. What does Linux have on par with QFS?
Also, Solaris doesn’t perform well on low-end single processor hardware. All of that scalability has come at a cost for Solaris on the low-end. The desktop/workstation experience for Solaris is considered by most to be terrible. The sound system makes the recently antiquated OSS from the 2.4 kernel look (and sound) impressive.
These are roles for which Solaris isn’t intended, and personally I’d rather Solaris remain throughput optimized than trying to optimize for the desktop experience like Linux has. Linux is known for having terrible I/O throughput in comparison to its Unix derived brethren and applying ideas to lower latency into its I/O schedulers such as round robin I/O queuing.
In general, Linux is being optimized as a desktop operating system and Solaris is being optimized for servers. Yes, while many aspects of this are compile time options, many explicit throughput/latency options (such as the I/O scheduler algorithms) are not configurable at compile time and are inherent design decisions made by the kernel developers which optimize for latency over throughput.
Personally, I would rather Solaris remain unoptimized as a desktop operating system and that it continue its slant towards massive throughput.
> There’s also cluster filesystems to consider. What does Linux have on par with QFS?
SGI CXFS and ADIC CVFS have been running on linux for the past one year.
>> In general, Linux is being optimized as a desktop operating system and Solaris is being optimized for servers.
So the recent 1024 CPU NCSA Altix linux server is a desktop ?
The linux kernel is very flexible and can be configured to run on any class of machines.
IRIX used to run happily on a 180 MHZ O2 as well as a 1024P
supercomputer.
Quite similarly,linux runs fine on my 200 MHZ laptop and on
our 1024P systems here.
Please stop spreading FUD around like this.
I am seeing quite a few misinformed comments which claim that linux doesnt scale well.
So spreading misinformation yourself will make it better?
Please take a look at Altix where linux runs happily on 512 CPUs without any trouble. Linux is also going to run on the NASA 10000 CPU Project Columbia machine.
These are not running stock Linux kernels, and make use of bolted on third party code that is nearly as alien to the Linux kernel as a Windows driver would be. Solaris implements everything it needs to scale to really high end machines as part of the kernel’s basic architechture – it’s second nature.
I am not sure why people would still want to keep disparaging linux wrongly like this.
I just don’t like people to live deluded lives any more than I like to live in ignorance myself. For crying out load, it’s just software, and I’ve no issues speaking of it’s flaws. Grow up.
I am seeing quite a few misinformed comments which claim that linux doesnt scale well.
Please take a look at Altix where linux runs happily on 512 CPUs without any trouble.
Linux is also going to run on the NASA 10000 CPU Project Columbia machine.
I am not sure why people would still want to keep disparaging linux wrongly like this.
I think you’re misinterpreting the ability to scale to managing a larger number of competing threads, or greater demand on the VMM/VFS/network stack versus the ability for a single system image to run on multiple CPUs.
Yes, Altix supports an impressive number of CPUs in a single system image, but that’s more a limitation of the underlying hardware than the operating system.
All high end Sun servers utilize a crossbar architecture whereby there exists one point-to-point bus linking every card in high end SunFires for each slot in the system. This assures that there are no I/O bottlenecks in maintaining cache coherency.
I have no hands on experience with Altix, but it’s my understanding that the number of links per node is not equal to the number of nodes, and that I/O contentions do arise when trying to maintain cache coherency. While this isn’t a problem for scientific computing applications, it would certainly be a problem for enterprise databases.
Also in order for Linux to run on Altix it requires considerable modification by SGI… in other words, it’s not really “Linux” as exists on kernel.org but rather a substantially modified kernel based on Linux. As SGI’s future seems somewhat grim at the moment anyone purchasing an Altix system needs to face the possibility that SGI might not be there one day to provide updated, patched kernels, and users may be left with an enterprise caliber system from which management expects enterprise caliber stability and they must either cope with any problems in the existing kernel or attempt to roll their own from SGI’s patches. For high availability enterprise systems, neither of these seems a particularly attractive option… either be bereft of updates and support or ride by the seat of your pants with no testing except what you do yourself.
Ultimately, especially in the enterprise environment, “proprietary” isn’t a bad word, because with a proprietor comes testing and support. We see SGI offering much the same thing with Altix that Sun does with SunFire, and with it the same benefits and problems of any proprietary Unix operating system. The only difference is that while Sun’s future looks grim, SGI’s looks grimmer. SGI’s market cap has dropped from $700 million in February to almost $400 million now. I’d say they’re fairly ripe for a takeover…
…you should know all this, you work at SGI!!!
So the recent 1024 CPU NCSA Altix linux server is a desktop ?
Altix seems to have made great in roads in the scientific computing market. However systems applicable to scientific computing are certainly not comparable to systems intended for enterprise databases.
The linux kernel is very flexible and can be configured to run on any class of machines.
Do you have any performance benchmarks of enterprise databases, such as Oracle, on the Altix architecture? CORBA enabled financial applications? Anything to serve the needs of enterprise and not scientific computing?
Scientific computing models are typically enormous nested loops, and consequently spend most of their time doing computations on the CPU as opposed to making system calls and thus using kernel services. For example in benchmarks I performed on our atmospheric modelling application, its time spent in process context (user) was well over 400,000 seconds, whereas the total amount of time spent in kernel context was about 100 seconds.
Scientific computing applications do not often tax kernel services, whereas enterprise databases and other enterprise applications do.
Please stop spreading FUD around like this.
I gave specific examples of explicit latency/throughput tradeoffs which the Linux kernel developers have made which cannot be altered in any other way than rewriting kernel subsystems yourself. Please read and respond to the facts in my comments rather than declaring them FUD because you do not agree with my overall opinion.
The sound system makes the recently antiquated OSS from the 2.4 kernel look (and sound) impressive.
Oh puhleeze!!!. You should try our commercial OSS drivers – hey it’s free beer now – test it out at http://www.opensound.com/download.cgi
Just talk to some audio developers and they will tell you that the API is just too complicated – over engineered under delivered. Why do you think most apps are still using OSS API? You think ALSA’s so great, good, disable the OSS API support and good luck getting anything done in Linux but don’t go spewing mis-information about “recently antiquated” OSS API. You want proof……email me directly.
We’re coming out with OSS V4 and it will certainly beat ALSA in performance, features and ease of use.
OSS on Solaris works exactly the same as our Linux version.
All the OSS API compatible apps from Linux – our XMMS or Mplayer or Xine or Ogle work today with OSS/Solaris. Apps compiled to ALSA 1.0 will not work with ALSA 1.0.6 – so much for backward compatibility or stability of the API (and ALSA’s been around since 1996 – a year after we started OSS). But if doing cvs recompiles all day long is your thing, I certainly don’t have anything to offer you. OSS just works (like XMMS).
best regards
Dev Mazumdar
It runs Java applications faster while using less memory.
http://www.computerworld.com/developmenttopics/development/java/sto…
Now come on, somebody step up and tell me that Sun’s flagship OSes ability to run Sun’s flagship software platform is irrelevant.
Now come on, somebody step up and tell me that Sun’s flagship OSes ability to run Sun’s flagship software platform is irrelevant.
With the world split between writing either POSIX compatible apps or .NET based ones, the fact that Sun’s flagship OSes ability to run Sun’s flagship software platform is irrelevant.
With the world split between writing either POSIX compatible apps or .NET based ones, the fact that Sun’s flagship OSes ability to run Sun’s flagship software platform is irrelevant.
The Sun Messaging and Calendar servers are Exchange compatible, and one of the only real practical alternatives to Exchange. Interoperability can occur higher than the API level.
Ever notice that this comes up on every discussion that could possibly involve Linux. Fud fud fud cried the linux users. I for one grow tired of hearing it. But that’s just me.
Why is the first comment posted to an article like this always, “Gee linux could use (insert feature here)” or “Well it won’t be GPL like linux”? I really think that people need to get over it. Sun has been producing a world class operating system at the expense of many engineers and dollars for many years. It’s their code, they paid for it. It’s great that they’ve decided to let the world look at it and comment. It’ll be wonderful for actual customers to be able to repair issues at source level if they have the resources, or have others fix it for them in an open manner. I personally DO NOT want the best features of Solaris stolen and put into Linux. At that point Solaris loses all value and will be killed. My personal hope is that they use the Apple Darwin model where the core is open and the sun specific stuff stays with Sun. Whereas in a perfect world the entire system would be open, in the real practice of things, software needs to retain value in order to remain viable. Sun is a hardware company just like Apple, and if their software does not hold it’s own value, their hardware business will suffer.If that happens, it’s one less company for Linux to mimic. I am constantly reading how Linux will kill Unix. If the Linux crowd comes up with an insanely great new implementation of an operating system I will be pleased, but as long as they are just copying previous designs I am not impressed. As it is now you still get what you pay for. Oh, and to stay within the scope of the article, Solaris 10 blows away anything I have seen from Sun in the past.
And so is anything released under an OSI approved license, which Sun has already said Solaris will be.
Not quite. Open Source Software and Free Software are different things.
Well, I will “step up” and mention that the word “Linux” is not even in the text of the article you linked.
How this article proves that Linux has something “on” Solaris is beyond me, much less that it can “run Java applications faster while using less memory”.
Oh puhleeze!!!. You should try our commercial OSS drivers
This is the point…out of the box, the open source version of OSS (yes I include the drivers as part of this) is inferior to ALSA, this is nearly universally agreed upon.
Also, last I knew, Sun didn’t ship OSS commercial or any other flavor with Solaris. So I don’t even see how your comment is relevent. Sun’s built-in audio system is generally agreed upon to be pretty sucky. Also, not every Solaris-native app is written to support OSS.
Why do you think most apps are still using OSS API?
Actually, most apps are using higher level APIs such as GStreamer, ARtS (I think that’s the capitalization), ESD, or Jack. Applications like XMMS support output plugins. The writing’s on the wall…ALSA is the future in Linux-land.
We’re coming out with OSS V4 and it will certainly beat ALSA in performance, features and ease of use.
Good luck getting mainstream distros to include it. Honestly, good luck, I love seeing competition. Having just one version of anything sucks. (See XFree86 vs. X.org)
OSS on Solaris works exactly the same as our Linux version.
No, actually it doesn’t because OSS is included out of the box on some Linux distributions…you have to download and install it separately on Solaris.
—–
In short, under any sort of unbiased scrutiny, your arguments hold no water at all. There is nothing at all wrong with what I originally said.
Solaris builtin audio system < Linux OSS < Linux ALSA
“i don’t think there are many people out there who can improve such a complex os.”
If its written well, it won’t be impossible to improve. The Sun engineers work on it after all…they arn’t gods.
” but there are tons of 1337 h4x0r5 out there who need something to rip off.”
So? You really think I could just copy Solaris..rename it…Moonres and go sell it? Nobody would actually buy it and depending on Sun’s license I could get my ass sued off.
I suppose I could copy code out of it into my personal OS (assuming I had one), but it would be difficult. Porting the code would introduce possible bugs outside what was in solaris. It is unlikely that my OS’s design matches up perfectly with Solaris..so this would be pointless for me.
“for sun as a company i think the microsoft-way is better than oss.”
Why?
“ms gives you access to the win sources, but you have to sign some contracts for this.”
There are a huge number of devs who would never do that. So Sun would be in the same position they are in now..which isn’t working.
There are many “proprietary” zealots who defend any proprietary system against linux. “MacOS is better than linux”, “BeOS is better than Linux”, “OS/2 is better than linux”, “Solaris is better than windows”, blablabla.
And there are some crazy people who say “Windows is better than linux”.
Solaris on Intel is known as “Slowlaris”. CDE is horrible and Sun is at trouble with linux pressure.
Novel also was king and nowadays they are making … linux.
Unfortunately M$ is more difficult to win because of “chicken and egg” problem of offer of commercial applications.
“Why not just use Solaris?”
Becuase Linux is free. And because GNU tools are much better than Solaris, or and other UNIX (with the exception of AIX maybe). And because all Linux is missing to put UNIX to rest is some of UNIX’s scalability.
UNIX scalability + GNU tools + 2.6 == The shizzle
Cut the crap, you can get all the GNU crap on Solaris without the need to butcher Solaris into some crappy opensourced, Stallman enspired wet dream.
Want gtar? then use it, its located in /usr/sfw, want to use KDE or GNOME? download it off blastwave.org and install it. Everything that is available on Linux is available on Solaris.
looks like someone doesn’t like all those comments on the first page.
<OT>
That 8-core CPU sure looks tasty. I’ll take a couple of those running at 3 Ghz please. That might satisfy my xvid/theora encoding needs for now.
I doubt I’ll ever see Solaris running on my desktop.
But this is what I call competition. Good job Sun exec snobs. I’d suggest toning down the IBM bashing commentary and raising the volume on your technology and roadmap. Where do you want to take us? Or is your plan to drag us kicking and screaming?
There are many “proprietary” zealots
Nope, sorry the zealotry is on the linux side. The rest of us really don’t care that much about software. After all it’s just that, software.
I am seeing quite a few misinformed comments which claim that linux doesnt scale well.
Please take a look at Altix where linux runs happily on 512 CPUs without any trouble.
Linux is also going to run on the NASA 10000 CPU Project Columbia machine.
I am not sure why people would still want to keep disparaging linux wrongly like this.
Please, just because the guts of the linux kernel detect 512 processors, doesn’t mean that the parts that make up the operating system will be able to adequately take advantage of those 512 processors.
Solaris is fine grained from the top down, tweaked, modified, and tuned for ultra high scalability, every part of the operating system has been ultra-fine grained, and deadlocks are a rare even in the life of Solaris vs. the weekly race conditions that are constantly found in Linux.
Hey, that doesn’t include the number of scalability issues with file systems, constantly VMM problems, driver conflicts between the the new scheduler and parts of the kernel.
I think Linux is killing Solaris. 5 years ago, would you ever believe that SUN would ever fathom open sourcing Solaris? I fail to understand what SUN intends to acheive by open sourcing Solaris.
I also don’t think Solaris supports as many hardware devices and filesystem as Linux does. It is not as customizable as Linux is, neither is it as portable. So what’s the point?
Linux winning point today isn’t its features. It is the fact that it is ridiculously flexible, can be a customized to fit your business model and problem and can run accross different hardware platforms.
The fact that Linux is free was an early advantage. Today, it’s just a side bonus. Conclusively, Linux’ customizability, flexibility and portability is second to none. To make matters worse for Solaris, Linux it is free. How can anything compete with that?
And by free, I mean GNU free.
Solaris has traditionally run on low MHz multiple processor systems correct?
I had the thought that now with AMD and Intel releasing Dual Core/Multicore processors in the future, that Solaris is the best match with these system simply because there code in this situation is more mature.
In lectures at school we are taught that it is highly scalable and multithreaded. So Solaris on AMD/Intel will have the parallel speed and also the Ghz.
http://arstechnica.com/news/posts/20040912-4176.html
And by free, I mean GNU free.
There are better things in this life. Like BSD free.
The fact that Linux is free was an early advantage. Today, it’s just a side bonus. Conclusively, Linux’ customizability, flexibility and portability is second to none. To make matters worse for Solaris, Linux it is free. How can anything compete with that?
By being both free *and* technologically superior? Yeah, that’s right, such things exist. As far as stability, scalability, and secureity are concerned, Solaris beats Linux hands down. The BSDs are holding their own as well. Linux just isn’t the be all, and end all of software, open source or not.
I also don’t think Solaris supports as many hardware devices and filesystem as Linux does. It is not as customizable as Linux is, neither is it as portable. So what’s the point?
Like I said before, there are better things. BSD has been ported to far more platforms than Linux. BSD is also more secure than any Linux that’s been cobbled together so far. Not to mention the fact that Linux isn’t *that* much more scalable than BSD.
I think Linux is killing Solaris. 5 years ago, would you ever believe that SUN would ever fathom open sourcing Solaris? I fail to understand what SUN intends to acheive by open sourcing Solaris.
I think that 5 years ago, Linux killed your ability to think logically.
@Kaiwai: I love how you use “opensourced” is a pejorative. All else being equal, an open-source program is better than a closed-source program. And what did Stallman ever do to you? It is because of people like Stallman that UNIX on the desktop is even the remotest possibility. The only thing the commercial *NIX vendors were doing in that space until Linux came around was forking it over to Microsoft. The only software that makes Solaris even remotely palatable on the client side is crappy opensource Stallman-inspired software like GNOME.
Also, your comments about Linux’s scalability are off-base. SGI ships huge Altix machines with Linux. The Altix has demonstrated very good performance in testing. Clearly, Linux’s scalability goes beyond just detecting the 512 CPUs in the system. Lot’s of people are using Linux on some very big machines. Simple zealotry doesn’t get people to buy 512 CPU Altix machines. There has to be some merit to the system.
@Jared: I think you’ll find quite a bit of zealotry on both sides. I’ll refrain from naming names, but perceptive readers will figure it out.
I wonder how many security flaws in Solaris are going to be discovered when everyone can look at it.
I do like Solaris, though. For a commercial Unix, it’s very nice, especially with Gnome running on top. The x86 support seems to be much improved now, too.
With regards to security, I doubt you can say plainly that BSD (which BSD?) is more secure than *any* Linux. I’m sure NSA’s Linux is pretty secure. Also, none of the free BSDs are scalable as Linux. FreeBSD and DragonFly BSD may get there, but they are not near there yet.
With regards to security, I doubt you can say plainly that BSD (which BSD?) is more secure than *any* Linux.
Secure OS. BSDi based, not one exploit. Ever.
Secure OS is owned by the folks that developed the technology that the NSA’s SELinux is based on.
As Bender would say,
“Bite my shiny metal ass.”
Also, your comments about Linux’s scalability are off-base. SGI ships huge Altix machines with Linux. The Altix has demonstrated very good performance in testing. Clearly, Linux’s scalability goes beyond just detecting the 512 CPUs in the system. Lot’s of people are using Linux on some very big machines. Simple zealotry doesn’t get people to buy 512 CPU Altix machines. There has to be some merit to the system.
OK, reluctantly Ill rehash what has been said before, because it just doesnt seem to be getting through. Altix HW/SW is made and bought for scientific workloads–tasks that generally spend very little time in the kernel and are often embarassingly parallel. There is a reason that you never see Altix benchmarked with something like Oracle, even though it would seem to be an easy thing to do. On commercial workloads, where a lot of cache coherency checking happens and IO throughput is crucial, the Linux / Altix combo would not impress.
Still, it is fair to say that Linux does better than just detecting 512 CPUs. Whether it is suited to running anything more than processor-bound tasks on these CPUs is another question.
OT, and I know this is just throwing fuel on the fire, it may be instructional to look at the top performance non-clustered TPC-C and TPC-H results on http://www.tpc.org . On their Itanium 64-way TPC-C submissions, HP uses HP-UX and Windows Server 2003, and no one is using Linux on any of the top TPC-H submissions.
Linux is a cool OS. But there are other cool OSes out there.
Well, it’s impressive that SecureOS has never had an exploit, but it also seems to be a rather niche product. I’m sure it’s very secure, but I don’t think you can present any evidence to support your catagorical statement that “BSD is more secure than any Linux that has been cobbled together so far.” If you’ve got a study comparing SecureOS with NSA Linux, I’d love to see it. Otherwise, you’re just making speculative comments.
PS> And why? Because you Mr. Clueless Zealot, bother me.
Huh? That doesn’t make any sense in the context of any post I made. Are you responding to me?
According to the site you linked TPC-C spots #8 and #10 for non-clustered cases are held by Linux machines. The top for all machines is held by a Linux cluster as is spot #9 in the overall test. Who fills in the majority of the top of the TPC-C test? AIX, not Solaris. So much for Solaris being the creme de la creme of OS’s for this task. Solaris fairs better in the TPC-H test, but Linux is again the top OS for the overall results for all of the cases except the 10,000GB case, which is held by AIX not Solaris (again). So it sounds like Linux can hold its own in this set of benchmarks. It helps that now you have IBM starting to push their high performance Linux initiative.
On the guy’s blog entry…really cool processor but this guy has unfortunately (and obviously) actually started believing their own PR spin. That’s sad.
Well, it’s impressive that SecureOS has never had an exploit, but it also seems to be a rather niche product.
All relative wouldn’t you say? Linux’s ~1% of the desktop market seems to be a rather niche product eh?
I’m sure it’s very secure, but I don’t think you can present any evidence to support your catagorical statement that “BSD is more secure than any Linux that has been cobbled together so far.”
You mean like what you’ve presented in your case against it? Your one solid example (SELinux) was pretty weak in the face of what I posted mere minutes later.
If you’ve got a study comparing SecureOS with NSA Linux, I’d love to see it. Otherwise, you’re just making speculative comments.
http://www.securecomputing.com/news_display.cfm?nid=466&lang=en
This is not quite what you’re looking for, and admittedly, it’s for their firewall appliance, but if you read the whole thing, you’ll notice that the certification that they’re going on about, includes Secure OS as well. I am tired, and I’ve lost the ability to google, but there’s also a similar news bit about just Secure OS getting that certification. Has *any* SELinux distribution, including that provided by the NSA themselves done the same? Have they even come close?
Look also to any large repository of OS compromises. You’ll find plenty of exploits for Linux (pick one, google yourself, they’re a dime a dozen), and not one for Secure OS. I am not making this stuff up, and my patience for people who refuse to do even the most basic bit of research for themselves is wearing thin.
And WRT portability, don’t make me provide links to NetBSD’s web site. Linux doesn’t even come close to the kind of portability that these guys have managed to accomplish.
And those benchmarks that I’m sure you were refering to that showed Linux to be the king of OSS OS scalability also showed (if you looked for more than a few seconds) that it was not a tremendous gap seperating the three best ranking.
At any rate, it’s late. I’m tired. You should read up on this stuff before posting again. You never know. You might learn something.
With regards to TCP benchmarks, interested readers should actually take a look:
TCP-C: http://www.tpc.org/tpcc/results/tpcc_perf_results.asp?resulttype=al…
TCP-H:
http://www.tpc.org/tpch/results/tpch_perf_results.asp
A Linux cluster is #1 in TPC-C, and has two #1 spots (out of five) on the TPC-H benchmarks. If you discount clustered systems, there are two Linux machines in the top ten in TPC-C. Interestingly, #9 on TPC-C is a 32-way 1.5GHz Itanium2 NEC machine that delivers almost the same performance as the 64-way 1.5GHz Itanium2 Superdome you mention.
Also, I think you underestimate the I/O requirements of scientific computing. There is a reason the NASA 10,000 CPU Altix cluster will have 500 terabytes of storage.
Anyway, it is clear that Linux is very scalable. As scalable as Solaris? Maybe not, but I never said anything of the sort. If you inferred that from my statment, read it again until you get it. My response was to the idiotic crack about Linux just detecting the 512 CPUs in the Altix.
According to the site you linked TPC-C spots #8 and #10 for non-clustered cases are held by Linux machines. The top for all machines is held by a Linux cluster as is spot #9 in the overall test. Who fills in the majority of the top of the TPC-C test? AIX, not Solaris. So much for Solaris being the creme de la creme of OS’s for this task. Solaris fairs better in the TPC-H test, but Linux is again the top OS for the overall results for all of the cases except the 10,000GB case, which is held by AIX not Solaris (again). So it sounds like Linux can hold its own in this set of benchmarks. It helps that now you have IBM starting to push their high performance Linux initiative.
Sun does not “do” TPC-C. Probably because they can’t take the top. Which kind of indicates the value of being #8 or #10.
As for TPC-H, we are talking about different things. You are talking about clustered+non-clustered TPC-H results, I am talking about non-clustered. Considering that the original discussion was about Linux’s single system image scaling on very large systems, how it does on a cluster of 2-way Opteron systems is mostly irrelevant.
As far as for Solaris being the creme de la creme of OS’s for this task, I never said a word about that in that post. Didnt even use the word “Solaris”. My only point was that for big commercial workloads, Linux doesnt rule the roost.
All relative wouldn’t you say? Linux’s ~1% of the desktop market seems to be a rather niche product eh?
I’m not using “niche product” as a pejorative here. My point is that there is a fundemental difference between being a mainstream product (something that runs on millions of desktops, and a large percentage of servers), and saying you’ve never had an exploit, and being a firewall OS and saying you’ve never had an exploit.
You mean like what you’ve presented in your case against it?
I don’t have to present a case. I didn’t make the catagorical claim that “BSD is more secure than any Linux that has been cobbled together so far.” The burden of proof rests on you, not me.
This is not quite what you’re looking for,
No, not it’s not. That isn’t a study, it’s a marketing piece. Hint: a proper study would have a URL that didn’t start with www.company-in-question.com.
Look also to any large repository of OS compromises. You’ll find plenty of exploits for Linux (pick one, google yourself, they’re a dime a dozen), and not one for Secure OS.
I hope you realize how stupid this logic is. Search for BeOS exploits. You won’t find many of these either. That must mean BeOS is the most secure OS ever!
I am not making this stuff up, and my patience for people who refuse to do even the most basic bit of research for themselves is wearing thin.
Let me reiterate. You’re the one who made the claim: “BSD is more secure than any Linux that has been cobbled together so far.” It’s not my job to prove your idiotic claims for you.
And WRT portability
I never made any comment about portability. But since you bring it up, if you take out duplicates (big-endian vs little-endian), both have about 12 supported CPU architectures. I’m not going to try to get a comprehensive list (only Linux runs on my iPod, but only NetBSD runs on my EM500 PDA), but suffice it to say, both are highly portable, perhaps equally so.
And those benchmarks that I’m sure you were refering to that showed Linux to be the king of OSS OS scalability also showed (if you looked for more than a few seconds) that it was not a tremendous gap seperating the three best ranking.
What benchmarks are you talking about? If you’re talking about the ones that were published on OSNews, I wouldn’t put any stock in those. They say NetBSD scales better than FreeBSD! In any case, that’s not the kind of scalability I’m talking about. I’m talking about scalability on SMP machines. FreeBSD 5 is on it’s way there, but for god’s sake, FreeBSD 5 hasn’t even made a stable release yet.
Also, I think you underestimate the I/O requirements of scientific computing. There is a reason the NASA 10,000 CPU Altix cluster will have 500 terabytes of storage.
Admittantly, I should have been more clear about that. However, 500 TB is not so much for 10K CPU’s. The p690 at the top of TPC-C non-clustered has 74 TB for 32 CPUs, while the Integrity at #2 has 38 TB for 64 CPUs. At same the ratio of disks to CPUs on the Integrity, the Altix cluster would have 6 petabytes. The major reason for this is the need for a huge concurrent number of random seeks.
Sorry if I overgeneralized your post, I am just very tired of the old Altix Rulez All saw. Though it is a neat machine.
I hate to post and run, but us Louisianans are going to be evacuating for hurricane Ivan real soon now. Hopefully I can get back to you soon.
Where did I claim that Linux “ruled the roost?” I’ll freely admit that Solaris is probably more scalable than Linux on > 16-way machines. I’m just claiming that it’s hardly as bad as some people had claimed.
I also find this comment odd:
Sun does not “do” TPC-C. Probably because they can’t take the top. Which kind of indicates the value of being #8 or #10.
Okay, so #1 doesn’t count, because it’s a cluster, and #8 and #10 don’t count because they are too low. Way to set up the question to get the answer you want. I also fail to see how Sun not being able to get good scores in TCP-C reflects badly on Linux’s scalability.
Okay, so #1 doesn’t count, because it’s a cluster, and #8 and #10 don’t count because they are too low. Way to set up the question to get the answer you want.
I do not believe that the performance of a cluster of small machines can tell us a whole lot about the performance of a very large single image system just because they run the same OS. If you disagree, tell me why you do.
As for #8 and #10, of course they count. I just wouldn’t be so proud of it
~
I also fail to see how Sun not being able to get good scores in TCP-C reflects badly on Linux’s scalability.
The point was that just about every big vendor can get a 700,000 score. Most just dont bother to submit. Sun is either in that camp (likely) or they just cant be bothered, period.
It will be interesting to see what sort of license Sun comes up with. I wonder if it will open the door for affordable Opteron workstations from other vendors. Sun has come out with some decent low end Opteron servers, but their workstation line has suffered from poor price/performance ratios for ages. Solaris used to be the standard platform for engineering applications, but Red Hat has gained significant support from customers seeking better hardware value. Some may ask why Sun would be willing to give away hardware sales, but I think that they already lost those sales and would do well to keep customers and ISVs on Solaris. When all of the workstations have switched to Linux, switching the servers to Linux becomes more attractive.
On a minor side note, every time Sun starts talking about Red Hat, my respect for Sun goes down. Am I the only one rubbed the wrong way by this? Schwartz seems to have a special knack for making me distrust Sun management and raise doubts about Sun’s future. I have far more confidence in Sun’s engineers than in their upper management.
Please just download solaris express and try it out. First, be sure to install everything (entire installation), and check out the man page for “ata” if you’re using ide or whatever appropriate driver for a scsi controller (use prtconf -D) in order to tweak just a couple of settings in the driver’s .conf file. Second, select the appropriate XF86 driver for your video card in kdmconfig. I’m sure you’ll find that the “solaris” myth is no more.
As for who benefits from who when Solaris goes open source (provided that the license is GPL compatible), I think that for purely technical reasons, Solaris actually has the better opportunity here. Linux internals and its subsystems too tuned for specific tasks to easily port technologies from Solaris over. On the other hand, Solaris has much in the way of doing everything more robustly, if not leaning towards being a little too complicated, with frameworks that are **much more generic**. It is for this reason that in some cases Linux performs faster, but also the reason why many things can be taken from Linux into Solaris with less work. i.e. it’s easier to implement a thin translation layer in Solaris to mimick the services and functionality that Linux provides than the other way around. Just look at the Solaris’ STREAMS framework vs. Linux BSD socket network implementation. In Linux, it’s only for networking (just about). But in Solaris, as was intended in the original SysV/Dennis Ritchie/AT&T design, you can use it for anything, literally. So not only is it used in Solaris for network drivers and layers of the network stack, but also for terminal emulation, serial devices, console, keyboard, mouse, USB devices (audio, storagae, etc). etc. with services provided for multiplexing and asynchronous operation.
Sorry, correction: “”solaris” myth is no more. ” –> “”slow-laris” myth is no more.” With Solaris 10 the desktop performance (and overall i think) has noticably improved from previous incarnations.
I like this idea. After Solaris will be open source Linux will be history!!
Just some opinions:
“Becuase Linux is free”
…If your time is worthless (Rich Teer)
And, Linux is more expensive than Solaris if you buy from
Red Hat with Support.
“And because GNU tools are much better than Solaris”
Solaris has been shippipng with GNU tools since Solaris8
also there is blastwave.org, sunfreeware etc. Anyone tried
pkg-get? That utility rocks.
“Also, Solaris doesn’t perform well on low-end single processor hardware.”
Are you making this comparison based on Linux-x86 to
Solaris-SPARC? Solaris performs just fine on the deskstop,
it’s SPARC that’s not great running desktop apps. A lot
of the perceived benefits of Linux are actually those of
x86. Try Solaris10 on the same hardware.
TPC-C:
Sun does not submit results on this because the benchmark
itself is flawed and open to abuse. It’s very
partitionable and can be overoptimized. I think there is
something like 11 tables and about 5 functions :- it’s
not realistic.
And the Red Hat comment, I think Schwarz is trying to
say that in the markets that Sun is interested in Linux
= Red Hat. Customers are being sold into “Openness” with
Red Hat on the back of linux the social momvement. The
reality is they are just as locked in as if they had gone
to MS/IBM/HP/Sun etc.
> On a minor side note, every time Sun starts talking about Red Hat, my respect for Sun goes down.
I actually sort of agree with Schwartz in regards to Red Hat. Red Hat is really trying to hijack the Linux movement and lock as many Linux users as possible to its own distribution, they even started quoting the dollar figures of how expensive it became to migrate off RedHat on to other distributions (around $40 million) — it just irks me. And this is coming from the the company whose biggest contribution to the technology behind Linux OS is the package manager. If anything Sun deserves more praise from Linux pundits than RH.
Open Source Solaris doesn’t mean that Linux loses. It means that everyone wins.
That whole zero-sum game theory idea that you can’t win unless someone else loses only applies to closed source, and not always there, either.
Free and Open Source means friendly competition, where ideas and improvements can be shared. Friendly, but sometimes noisy, like a bunch of kids playing. It’s also usually the fans on the sidelines who get unruly, not the players.
It’s great that there are many Linux distributions. Each drives the others to improve. We all win. It’s great that we have several BSDs. Each is better for some task, and they all learn from each other. It’s great that we have KDE and GNOME. Remember that unification of effort gave us CDE, but competition brought us both KDE and GNOME. It’s great that we have emacs and vi.
I don’t worry when we have too many projects solving similar tasks, I worry when we have too few. The technical term for a situation where there is only one solution is ‘stagnation’.
I don’t see lockin from Red Hat. Everything that they write is under a free license, so they can’t lock anyone in with that. They charge for support, as they should. If I don’t want the support, I can download the source and compile it.
I can also switch distributions. Surely moving from Red Hat to another Linux distribution is no worse than switching between any other two Linux/unix flavors?
Red Hat also contributes quite a bit, from gcc to GNOME to the kernel. The idea that they leach off of other developers without contributing anything sounds like FUD to me. I don’t see how Red Hat contributes any less than SUSE, and the two of them are well ahead of other commercial distributors.
The transition to Fedora was handled poorly, and irritated many, myself included. I’ve decided that the RH products are too conservative for my personal needs, and that Fedora is too unstable. I’ve moved on, and Red Hat no longer gets my support dollars. But missteps and all, I still think that Red Hat is a fine company.
lot of fun here
thanks guys