Slashdot Mirror
Sun Solaris 9 for x86 for Evaluation
Rune Tønnesen writes "Sun has listent to their costomers, they have a released Sun Solaris 9 x86 for test and evaluation purposes, it can be downloaded ($20) as part of their OE Customer Early Access software.""
I found Solaris/X86 (2.6->8.0) pretty stable.
In fact, it has been rock solid.
More so than Linux (Mandrake 8.2) on the very same hardware (serverworks mobos).
I spoke to someone at sun about 2 years ago who said that they'd like to OS solaris but some of it is still based on code that they licensed years ago from companies which are no longer around.
I'm not sure if this is still the case, but it'd still require quite a lot of effort to replace this code with cleanly implemented open-sourceable code.
It'd probably make reasonably good economic sense since x86 boxes just cant compete with the higher end sun machines - either in performance or reliability.
Cheap? You can purchase SparcStation 10's or Ultra 1's on ebay for ~$20-$100 that will easily run Solaris 9. Now they won't be the most spritely, but if the issue is getting something up and going _cheaply_ can't beat it. Sure is a lot better than messing around with pc hardware that will make Solaris x86 happy. Just plop the cd's in and go. A great way to learn Solaris on the cheap.
Sun says Sol9 is Linux compatible. They also include many of our favorite Open Source apps, and many of those are Sun supported.
If we didn't need SPARC binary compatibility for some of the libraries we don't have source code for I could probably convince the Powers That Be to take a look at this at work, especially since I could build a dual CPU Athlon 2400+ development box for cheap. (I have one at home. Real MP 2400+ chips should be available later this month, saving you the nuisance of hacking XP series chips.) Being able to use the same GTK+/GNOME GUI source for both Linux and Solaris development is very, very interesting. Windows has probably already won where I work, but who knows?
LX50 runs a sun version of Linux, hence the 'LX'.
Maybe you should read the specs on the LX50?
Around Windows 2000, if i'm not mistaken?
You're not
Don't forget to think different.
Or you can just download it if you don't want support. [shrug].
May we never see th
IBM is in the middle somewhere; on one hand people buy IBM (among other reasons) because they know IBM thinks things out the first and their systems don't change so fast. You get a solid (usually) platform which doesn't have a lot of flux. On the other hand IBM is rapidly porting everything in AIX worth a crap, like their volume manager and their filesystem, to Linux. They're also working on support for excessively multiprocessor systems, right? So soon there will be no reason to run AIX on RS/6000 except legacy apps. If IBM is smart they'll produce (and sell) an AIX emulation package for Linux and phase it out over time, putting their effort into Linux. Then they can make a new release of IBM/Linux (hee hee) whenever there's a new minor stable kernel revision, and point patches thereafter. It might also make sense just to use the linux kernel and stick with all of their commands and utilities. In fact, now that I consider it, this seems the most likely long-term road for IBM.
Solaris is a pretty cool OS from a Unix standpoint, they certainly do things in the Unix way. It's a healthy SVR4 clone with plenty of added functionality. Sun's package manager was clearly designed from a Unix mindset. Their init system is classic System V. The system is easy to work on because it doesn't attempt to shatter your preconceptions about Unix; It looks like Unix, smells like Unix, works like Unix. The only real bummer is that you have to pay a whole hell of a lot for a compiler from Sun, or run GCC which has traditionally generated pretty slow code on sparcs. I guess GCC 3 is supposed to be MUCH better in that regard.
Sun makes most of their money with a) really nice hardware and b) really big service contracts for really nice hardware and peripheral systems. Selling those little 400MHz PCI Ultrasparc III PCs has got to be making them almost no money, but as long as you're not actually losing cash, increasing market share is always good.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
While getting training on Solaris is invaluable for any *nix sysadmin worth his/her salt, it's my belief that when it comes to experience helping secure a job getting that experience on x86 hardware lies somewhere between "next to useless" and "better than nothing" on the usefulness scale. Anyone that wants Solaris software experience will also want Sparc hardware experience (disk arrays, remote mgmt cards, sbus legacy stuff, etc -- things you don't normally see on commodity PCs). They'll probably want someone who knows enough "Sun" to know what the difference beween an E420 and a SunBlade is and won't get surprised to discover that one of them doesn't have anything more than a console attached to it.
If you want Solaris experience for a job, then you'd be better off buying an old Ultra 5 for 80 bucks than paying for beta x86 software. You'll at least be able to say during your interview that although you don't have any "real world" Sun experience, you have been playing with an old Ultra in your spare time in order to get up to speed or round out your professional experience. I've seen a few people get jobs this way in fact.
You have a much better chance if you get an old Sparc, stick it in the corner, hook up a serial cable to it and run BIND on it for internal DNS or something than playing with x86 Solaris on a PC.
-B
Ash and Hickory, straight-grained and true, make excellent bludgeons, dandy for the cudgeling of vegetarians.
Funny, it wasn't too long ago that the rumour mill was churning out stories about Sun buying Apple. Times do change, don't they?
Sun and Apple probably don't have a whole lot to offer each other as long as Sun stays on the Sparc platform and Apple on PPC. Both have install bases that are far too big to change over. Sun's workstation market doesn't need pretty boxes or built-in screens, and apple's server maket doesn't need 64-way SMP systems.
So right now I don't think they really have much to contribute to each other's tech, since Apple wouldn't want to lose sales of their own workstations by giving sun Quartz, and Sun wouldn't want to lose any of their server sales by giving apple access to their interconnect technology.
___
Cogito cogito, ergo cogito sum.
Staroffice was not written in Java, so I'm not sure where you get the impression it was meant to be a Java showcase.
Sun purchased the software from a German company at version 5 (the one you probably used 2 years ago). Version 6 is a big jump in usability and performance (even though launch speed is slow).
Is it MSO? No, but its much cheaper and it get the jobs done. Sound familiar?
If you want to look at a very similar product, go to www.openoffice.org and download the open source cousin of StarOffice. Its not great, but its not bad - and its free!
I seriously doubt that anyone is using X86 Solaris in a production environment. I strongly suspect that much like myself, many people downloaded the OS and slapped it on an old Dell or Compaq desktop in order to learn something about Solaris. I found it beneficial, I learned enough of Solaris to be comfortable with it. I am now a fan, and would not hesitate to employ Solaris in a production environment running on Sun hardware. Without the free download, I never would have become comfortable enough with Solaris to consider it as an alternative to Linux and BSD systems. I suppose some had the opposite reaction though, installing it on junk hardware and hating it.
Completely OT, but the Admintool is crashing because you are running XFree86 instead of Xsun and you have not put the xfs fonts into your FontPath.
Add those fonts by running
$ xset fp+ tcp/localhost:7100
Their market is small, their niche is narrow, and their execution is bush.
... and their enterprise computers are just about the best in the business. If you want a powerful mainframe type computer, Sun is your go-to guy. They offer superrior hardware (Ultra III risc chips, etc), great service, and a fantastic operating system. I've heard people claim that solaris sucks because it's so archaic. This is exactly why it works so well - it's been around for a long time, and it's tried and tested.
There's an age-old balance, people. It's called ease-of-use versus power-and-stability. Solaris is not easy to use. It's harder to use than linux. But compare solars 7 to linux... solaris scales well down to the 12 Mhz sun 4c IPC range, while the same OS works great for enterprise servers with 64 Ultra III 500Mhz risc chips. That's scalability. It wasn't until solaris 8 that they gave up on the 4c arch. The 4m still scales well (50-110 mhz range, etc).
If you're looking to buy a 1.3 million dollar computer, you look at sun. The small-computer market isn't the majority of their business dollars. It's the top dogs. Yet, they still listen to the people who like solaris enough to want to use it on x86. How can you fault them for this?
Granted, 1.3 million dollar computers make up a small "niche" of the market. But someone has to fill it, and there is a lot of cash in 1.3 million.
sig?
Agreed. Very few people run sun software on x86 hardware in a production environment. Most of it is for testing, and/or just checking it out to see what it's about.
Don't run solaris on x86 arch because it's supposedly better. On x86, there are much better OS's. If you really want an SVR4-ish nix, use a bsd or something. I don't know of any popular ones other than solaris. Whatever.
Solaris hardware - I don't know about it being stable - it is as much as anything else, but it lasts forever and is hard working as crap. We still use Sun IPC's at my job, they're 12 mhz, late 80's or early 90's I think, and they still work great. Some of them the batteries have gone out - imagine that - the (soldered in) cmos battery goes out before the motherboard/proc/ethernet controler, etc fail. They're great for console access - if they don't detect a keyboard and monitor, output straight out the serial port.
Solaris hardware doesn't run D.net fast, but it sure does compile things fast. Startelingly so. We have a Dual-Pentium III 1.4 Tualtin with 3 gigs of ram, and it compiles things significantly slower than our 4x300Mhz Ultra II with 1GB of ram, despite being "twice as fast".
This is the advantage of sun. The hardware rocks. The software is built to match the hardware. I think it was more of them saying "yeah, well, if you guys want it on x86, here you go, but be aware it sucks." It might have actually made them money in that people would buy sun hardware after trying it on x86 and giving up on the crappy hardware.
sig?
OK, the spelling is lousy, but the meaning is pretty clear. From the look of this guy's nick, he's scandinavian, probably Norwegian. Maybe you're a polyglot, but I'm not, and I have a lot of respect for people that have taken the trouble to learn my language (perhaps imperfectly), when I'm too lazy to learn theirs.
If you really want an SVR4-ish nix, use a bsd
Huh? That's like saying "If you really want an apple, have an orange."
-- Never hit a man with glasses. Hit him with a baseball bat.
The truth is that it doesn't scale down so well. If you can do the comparison on a small machine 2 or 4 cpus, you'll see that bsd or linux are faster. Slowlaris^WSolaris is really optimised for the big irons not the smaller machines.
Cool, that is nice to know. Personally, I just took advantage of the free download for Solaris 8 to learn a bit about the OS. As for the Dell or Compaq server comment, I have no doubt that they could run Solaris in a stable manner. I was actually considering people like myself who installed Solaris on a Compaq Deskpro or the Dell equivalent. For what it is worth, it runs about as well as FreeBSD or Debian on a 400MHz Compaq with 380MB of Ram. Not bad at all really. Cheers.
OK, the O/S on IPC/IPX range is Solaris 1.x. I don't think you can put 2.x on them at all. This is SunOS 4.x which is the old BSD-based version.
The O/S on more modern hardware from the 50MHz Sparc 10 to the Ultra III belongs to the Solaris 2.x series. Solaris 7,8 and 9 are really Solaris 2.7, 2.8 and 2.9 respectively.
I don't think Solaris 9 support the Sparc 10/20 series anymore.
It is still rather scalable, as you say.
Microsoft did challenge hackers to break into a Windows system, didn't they?
They did but the box was effectively DoS'ed by the participation so there was no benefit. Plus there was no cash incentive for people who found bugs/exploits.
So we are left to test the production OS. And just as it starts to mature, Microsoft drops support and releases another version. In 2005 when MS drops support for Win2K, it will likely be the most secure Windows OS available at that time. But then they shoot themselves in the foot and remove it from the product catalog. All for the love of money.
This is where the DOJ needs to intervene. It is too bad that MS owns the DOJ, unfortunately.
Life is the leading cause of death in America.
IRIX has a number of assets that Linux does not have, even in the kernel space -- including scalability (support for up to 512 CPUs, 512 GiB RAM), advanced file systems (XFS journaled file system, XVM volume management), advanced networking (Clustered XFS, SAN), standards compliance (POSIX, DII-COE, Trusted IRIX), and a Unix (BSD+SysV) heritage -- that place IRIX in a different league from Linux and *BSD. It's not that these features could not be added to Linux, but at this time Linux and IRIX have different target markets.
AIX and Solaris also have features not found in Linux, I'm just not sure why you singled IRIX out. Don't forget that SGI has also developed a reputation, particularly for high-performance systems and cutting-edge hardware.
That said, many Unix companies do seem to be adopting Linux to some extent. Who knows what SGI will do?
but linux isn't SVR4, it was coded from the ground up. BSD was sort of forked from SVRx, but had every single piece of code rewritten by the folks at berkley.
Don't learn Solaris on a PC.
Please. For your production systems, you will have to work with differing device names and the open boot PROM. The OBP is quite different from your average PC BIOS, and you can actually program it in FORTH. In fact, we ask senior sysadmins who claim to know Sun hardware to describe simple secrets of the trade such as changing the hostid using FORTH.
Get a SparcStation to learn this and other fun Sun-specific stuff. SEVM (also known as Veritas Volume Manager) and the DiskSuite are also only available on Solaris, AFAIK and you must know those tools. Getting Oracle to run on Solaris requires kernel modifications, so you better know that as well.
In short, get a SparcStation 10 or 20 and learn this platform the right way.
I can tell you differences between quarterly Solaris releases, so trust me on this.
Leonid
Leonid S. Knyshov
Find me on Quora
Not sure what planet you live on but Solaris x86 has always out performed Redhat on my hardware. My latest comparisons are based on desktop-oriented tasks and server-oriented tasks both on the same laptop and workstation. Granted what is lacking is the latest device drivers that Linux is so quick to get, but since I only use IDE drives and 1-year old video cards, that has not been an issue. The laptop is a Pentium 166 w/ Chips&Technologies that needed XFree86 build to use full 1024x768x16bits possible. The workstation is a P3/733 w/ Nvidia TNT 32MB and no extra build of X was required to get 1600x1200x24bits .
.
The desktop feel was clean and fast on both OSes when idle. When not idle, even doing the cheapest disk task, Redhat graphic response really slowed down. Any significant background operation would really hit Redhat8 bad with Netscape7 performing really poorly. On Solaris9 the negative effect was barely noticable. The worst background jobs to hit Redhat were large filesystem operations, with a good mix of node and data I/O. When building XFree86 on both OSes in the background, it really hurt Netscape7 performance on Redhat, and was not even noticeable on Solaris.
The effects were consistent on both the laptop and workstation. The laptop simply exaggerated the effects more.
Now for Java. No comparison of anykind. Hands down, Solaris9 smokes Linux when it comes to running Java apps. I tested thread-crazy real-world servers where threads are not just token objects but are live and kicking expected to produce results. Not only did Solaris launch the threads faster, it's sycnrhonization across threads was much more optimal. I could easily saturate Redhat with a lower workload and see 100% CPU, while exact same workload on Solaris was 40%. These threads have a high amount of sychronization going on, and was the single largest contributor to the performance gap. Bottom line, big stuff runs better on Solaris. When not running big stuff, there was simply no noticable diff.
Granted I don't need all this OpenGL stuff or gaming, so that might be where Redhat outshines Solaris. Also probably video playback too but for that I use WindowsXP
Here's stuff I built...
gcc 3.2.1 (bootstrapped from SMC Solaris8 pkg)
GNU* array of make,fileutils,sh-utils...
netscape7
XFree86 4.2.1
XFree86 4.2.99.2
top, lsof, sudo
windowmaker 0.80.2
cvs
what native binary packages?
jdk 1.4.1_01
Acrobat reader 4.05
so what is Solaris x86 missing? Honestly all it needs is a god community packaging effort. Something like *BSD ports system wherer you can
install prepackaged binaries to well known (opt) location, or build them yourself to same or well known location, all with auto-dep recursive binary packge grabs or builds, as required.
Give the community that, and Solaris x86 will become more popular. Not everyone has the desire to build stuff.
>This is one of those things that really gets me. If Sun was really worried about stability and security, they'd be
>giving it away for the masses to put through the ringer. Hell, they could even put up a few boxes on the internet
>for a "compromise the box and win a prize" type of test. The dollar value of different exploits could grow daily.
You dont' actually believe that this is Sun's only procedure to check for bugs, do you? You can't maintain the quality of a product as large as Solaris with random testing as your only or primary means of testing.
The place to start is large amounts of automated regression testing and the use of other types of test tools. (Look at effect of implementing regression tests in GCC, the effect of the NFS testing tools on open source NFS implementations, and the effect of the Stanford compiler on the Linux kernel.) Then, once you've got the code in reasonably good condition, you give it to partners with a disciplined and rigorous testing methodology, and equipment suitable for stress testing. (I doubt many useful boundary conditions effecting my site will be found on Celeron 333s.) Once its passed through those tests, then its time to turn it loose for more testing by end users. Random testing by end users will identify bugs beyond the ones already found and fixed previous to this release. However Sun would have to be run by fools if random testing by end users just prior to release was their only, or even primary, testing procedure.
It is also worth keeping in mind that most of the code for this release has already been proven in the Sparc releases. Sun's goal for Solaris X86 has always been that it be bug for bug compatible. Most of the source tree is the same code.
>Eventually, they'd have *proof* of the level of security available.
Actually, they would have no such thing. Lack of a known exploit does not constitute proof that no exploit is possible.