Slashdot Mirror
Solaris 10 to be Open Source
An anonymous reader writes "It looks as though Sun is going to open source their new Solaris 10 operating system. It seems to include eveything except some device drivers. They plan to model the Darwin and Fedora projects. Sounds very interesting."
Major commercial programs like Oracle, DB2, WebSphere MQ are supported on Solaris/sparc, but not Linux/sparc.
If you've got sparc hardware, x86 stuff is a downgrade path you don't want to follow.
Scalability, stability are the main reasons. There are also some cool features like DTrace that aren't available in Linux.
Err, that's easy:
It's faster (approx. 30% : Sun to challenge Linux to a benchmarking duel shortly with Solaris 10)
It has N1 Grid Containers
At $99 It's cheaper than any enterprise Linux distro.
It scales better.
*Even* More secure than Linux
It's standard
Solaris 10 runs RH Linux apps efficiently
etc. etc. etc.
Exactly what I am wondering. Solaris is a descendent from the ATT/System V branch of the UNIX(tm) tree, not the BSD branch. They license the UNIX, not own the copyrights. Wouldn't they need permission from SCO (or Novell? ) and possibly a whole bunch of other people/corps/entities to really Open Source this stuff? Feels like heat, still looks dark.......
Last announcement about this was proven false by Sun's own CEO statments..
This will be the saem way with this announcement..
Don't Tread on OpenSource
Sun bought an special license from SCO, that lets them do whatever they want.
PENAROL: Seras eterno como el tiempo y floreceras en cada primavera.
The Non-disclosure terms for any protocols that can interoperate with Microsoft's Client or Server software would seem to restrict a lot of functionality from being released under an open source license by Sun..
Accorinding to section 2.1 of the Unix trademark use document this is not the case.
I can use a initial capital letters if I wish. *Their* convention is wholely capital letters.
Certainly not.
Stick Men
There already is "other people's code". For Solaris 10 they've adopted net-snmp over their own proprietary SNMP (which never did work worth a damn).
If I had to choose between a Solaris install, or a Linux install, on it's own, with a live IP address, I'd choose Linux every time
A Solaris install on the Internet on its own would probably get rooted before the hour ran out. At least it would if you were to choose a full install.
I use solaris on most of my servers, but before entering production, you have to patch the hell out of it (last time I checked, the Solaris 8 patch cluster was like 50MB), install ssh, if needed, and close a bunch of services that are activated by default *and* reactivated upon patch application.
I usually play it safe and install ipfilter, just in case.
No sig
parallel server was actually available in oracle 7, and possibly before. i know i was a dba for ops on 7.3 using the non-integrated dlm. it wasn't that bad for the large data warehouse we used it for.
After D. Robbins left Gentoo he has spent much of his time consulting with Sun. It appears one of his key strategic recommendations was to open source the OS and then infuse Sun's installation/package management system w/ portage. We all know portage is based off of BSD ports (at least in concept). So clearly, Sun is hoping to send Solaris down the path that Gentoo and BSD have already been down. The path to oblivion. By devaluing their intellectual property they can write it off and use that as means to boost their profitability (like they did w/ the Microsoft settlement).
Slashdot... news reporting and commentary on par w/ CBS.
That is why you have a support contract ! Sunsupport will solve a problem like that.
Bigger deeper sigh......
......Um no they won't, they have been out 3 times and done nothing, so much for that support contract
It said "windows 98 or better" so I installed Linux
Oh wait, that's right, there is none.
I'd say that the time of recess between the general community being aware of a vulnerability and a workable patch being available is a pretty good measure. But, according to this article, In 1999, Red Hat's "at risk" time was half that of Sun's (presumably then-current versions of Solaris), and a third of Microsoft's (presumably Windows NT 4). And that's with all the stuff that's included in the RH distribution for which there aren't equivalents included in Windows or Solaris.
Of course, it would be interesting to get more up-to-date stats, or stats for distros that are touted by some as being more security-conscious (e.g. Debian, OpenBSD).
--
It's also the name of a Roman god. Just because Microsoft has put their filthy hands all over another word, it doesn't mean that they now own said word. My firewall's hostname is janus, but that doesn't mean Microsoft has anything to do with it.
QUOTE
If you've got sparc hardware, x86 stuff is a downgrade path you don't want to follow.
UNQUOTE
Unless you want to talk about cost. If your software only runs on Solaris and your customers are balking on buying because of the high cost of Sun servers, you certainly want to investigate porting to linux.
Solaris 1.x is SunOS 4.x, which is BSD-based.
Solaris 2.x is SunOS 5.x, which is System V-based.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
The original SunOS was a BSD derivative. When Sun created Solaris, they combinded SysV with the BSD-based SunOS to get their finished product. Since then, Sun engineers have heavily modified Solaris to the point that it may contain little or no actual AT&T SysV code. What SysV remains could probably be fairly easily replaced with modern code from one of the BSD projects.
What's the last version of Solaris x86 you used? Considering Sun sells a ton of AMD-based systems and provides full support for Solaris x86 on them, I'd say you're a bit out of touch.
Solaris isn't only about scalability. There are lots of reasons why you might want to run Solaris instead of Linux on x86 hardware.
For instance, Solaris does updates and fixes via patches instead of by upgrading packages. They do a lot of testing so that when you get the "recommended" cluster of packages, you are starting at a major release of the system and adding a specific, known set of patches to the system. These patches have been tested before; by installing that exact set, you know you are using a configuration that is just the same as what Sun QA people have tested.
Meanwhile, with Linux, updates are pushed out in a more informal fashion. The assumption in the Linux world is that it's basically best to always have the latest version of everything, which is an interesting assumption. A geek friend of mine showed me the souvenir video from his first skydive, and the marketing blurb they put on there to assure it was perfectly safe said "we use nothing but the latest equipment". My friend and I looked at each other and laughed -- nothing but the latest equipment? But you said it was SAFE! If stability is important to you, then you want to start with a known good configuration and make the least changes possible to fix the stuff that needs to be fixed. Then you are introduced the least possible unknown stuff. Sure, Sun has occasionally released a bad patch for Solaris over the years, but overall they do a pretty good job with this. You can still get patches for releases of Solaris that came out years ago.
Another reason to use Solaris is the kernel. I won't deny that the Linux kernel has a lot of good technology in it. It's really not bad, and there is a lot of active effort going into it from several places. BUT, from the point of view of how much effort it takes to do the administration required for the kernel, Linux isn't all that hot. In fact, it's kind of bad compared to Solaris. In Solaris, EVERYTHING is a module, and all the modules autoload (and unload) totally transparently. You don't have put any statements into a module configuration file to force your ethernet module to load; it just does when the system notices the hardware is there. Likewise with absolutely every other supported piece of hardware (and other kernel modules, like filesystems, etc., etc.). Sure, on Linux you can autoload some things, but it isn't the default, and it takes quite a bit of work to make a kernel that you can install on 100 machines on your network and it will just load whatever it needs depending on the machine's configuration. Unless you're quite good with Linux, you're likely to just say "screw it" and build a kernel that does only what you need for a specific configuration, because building a general kernel that works on everything is too painful.
Furthermore, on Solaris there are stable BINARY interfaces for kernel modules to use, so you never need to rebuild anything. If you install some hardware that needs an add-on driver, you can just install the binary kernel modules, and they will work with the kernel you already have installed. (And if your hardware supports it, you can install them without rebooting, and the system will just start using them. You can even inactivate the hardware, remove the drivers, and install updated drivers, then reactivate the hardware without rebooting, if you want to.)
Still on the subject of kernels, what about /dev? One of the very
nice things about Solaris is that they
have got /dev figured out
already. With Linux, you have a bit
of a mess because devfs
works OK, but it's been declared
obsolete (or deprecated) in some
distributions, and it's no longer
being maintained. (And did the
memory corruption crashes ever get
totally fixed?) Meanwhile, udev
is coming along to replace it, but
udev is not re
That said, sgi propack linux will not scale to 512 CPUs on a general purpose oracle/SAP workload. Those beasts run because the apps are highly tuned to the environment. Those are distributed compute apps that spend a lot of time in the application and threading library, and very little code in the operating system.
Linux, especially 2.6, has much better thread granularity than 2.2 did, but it's not as parallel as irix or solaris.
Solaris and irix both evolved slowly to run on those huge boxes. First you thread the vm code, then the scheduler, then the buffer cache, then the filesystems, then the scsi drivers, etc. (not a precision list, but the point is that linux has the first several steps toward massive scalability, but not every step that solaris or irix have taken).
...that they'll GPL their source, or use a license anywhere close to the GPL. From the article is sounds like they're going to 'share-source' their stuff in the Microsoft fashion, then use doublespeak to call it 'open source'.
I doubt anything they call 'open source' will legally be able to be used in Linux.
Max
My god carries a hammer. Your god died nailed to a tree. Any questions?
How exactly do they run out of colors?
Sounds like either an older graphics card (cg3?) or an X11 configuration where the default color model is pseudocolor rather than trucolor. There could be good reasons for that (eg heavy use of a graphic application that modifies the LUTs for highlighting or animation effects) but it can have psychedelic effects on windows belonging to a different application, depending on the hardware.
-- Alastair
If you are planning on using a 100 cpu monster to run Oracle, you are first going to have to get Oracle itself to scale. It will have to manage a single SGA that is N times larger than something more typical.
...and to top it all off you've got a greater level of redundancy.
Oracle doesn't have a great history in this area. I have firsthand experience with 9i on E15Ks and secondhand experience with 8i on E10Ks.
Now, once you've got the instance itself sorted out then you have to be concerned about excessive block level contention. If you app isn't parallelizable at this level Oracle will quickly fall apart as it tries to manage block level locks and transaction consistency. Oracle will deteriorate into table level locks past a certain point.
If your app can't be deconstructed into a number of fairly distinct threads, it's not going to scale well period. Oracle has the capability to even scale datawarehousing apps where working on entire tables at once is commonplace.
Now, once you get into "the interconnect" itself that's not really a problem. That's the nifty bit about 9i. They fixed that part so that inter-node block transfers are more efficient than disk IO.
RAC performance gets a real kick out of this.
Now top this all off by the fact that 12+ cpu boxes tend to be remarkably more expensive than the next step down. That's why people run clusters. They don't want to pay for 100 cpus worth of E15K kit.
A Pirate and a Puritan look the same on a balance sheet.
http://www.blastwave.org
Sun should be doing this themselves - the Solaris package format is inferior and automatic dependency resolution should be expected.
Do you even know anything about perl? -- AC Replying to Tom Christiansen post.
You can also download the ISOs for Solaris straight off Sun's website, if you don't want to pay for media.
Another thing to note is that the "License" is probably nothing more than a sheet of paper that says "You're legally allowed to run this software on that machine". Solaris itself has absolutely no enforcement of licensing.
If you put 8 CPUs in that SS1000, and installed Solaris 8, it would work just perfectly fine with all the CPUs. (Support for sun4d machines SS1000/SC2000 was dropped in Solaris 9)
The question is: will it be free or is it only "open".
That won't meate the OSI's definition of Open Source. However, remember that their tradmark application for "Open Source" fell through. So, as long as Sun makes the source code viewable, then they can call it "open source", it just won't be "OSI Open Source". And definately won't be "FSF Free".
Journaled file systems reduce disk write latency because new data (and metadata) can be written in one sequential write. Non-journaled file systems (like ext2) must move the disk head to multiple locations to update the data and then the metadata. For more info, John Ousterhout (creator of Tcl) has some good papers about journaling file systems, such as "The Design and Implementation of a Log-Structured File System (1991)".
cpeterso
Actually 'pkg-get' used by Blastwave is
essentually a modified version of app-get
that uses the standard svr4 packaging tools
underneath. The most significant improvement
would be in Solaris 10 where the svr4 package
database performance has been improved with
a binary database. If you install a complete
Solaris 10 distribution along with Blastwave
you get some 1558 packages which God knows
how many files your admining. Have to admit
the 'pkg-get' from Blastwave is cool.