Learn about Linux
Download Linux
Get Linux helpSpecial Offers
Feature: Unix
Cyber cynic: Solaris on Intel -- forget about it already
Share
Print
Commentsof Practical Technology -
Way, way back in 1993 when I ran a Unix feature for PC Magazine, I helped review the first version of Solaris for Intel. I noticed at the time that Solaris on Intel wasn't the equal of Solaris on SPARC.
And why was that? It's because Sun has always made money from its hardware not its software. In short, Solaris on Intel was always meant to be a teaser to give you a taste of Solaris so customers would buy a SPARC box that would be much more profitable for both Sun and its resellers.
By the early 2000s, though Linux was eating Solaris on Intel's lunch and its market had shrunk to the point where Solaris on Intel was no longer a profitable way to get people to move to SPARC.
Mind you, it was popular. Sun's Graham Lovell, director of Solaris product marketing, has told me that more than 1.2 million copies of Solaris under the Free Solaris program have been downloaded, and that "the vast majority -- approximately a million -- has been Solaris 8 on Intel." Of course, Linux downloads are in the tens of millions and, in any case, popularity alone doesn't pay the bills.
So it was that early this year, Sun announced that there would be no Solaris 9 for Intel, in effect, killing Solaris on Intel's future. That should have been that. Instead, it turned out that there was a vocal group of Solaris on Intel fans. This group, Save-Solaris-x86 and Sun's been seemingly waffling ever since on whether they'll ever release a new Solaris on Intel.
Let's get real. Sun's not going to do it. There's no financial reward for them even if it gets if it makes a tiny, albeit very loud, group of Solaris on Intel fans happy.
The group's very arguments underline that they're not thinking about technology. They repeat the never publicly acknowledged strategy of Solaris on Intel. I quote from their San Jose Mercury News ad: "We believe that the 32-bit x86 architecture can never be a true competitor to Sun's 64 bit SPARC. In fact, Solaris on the x86 platform complements Sun's SPARC solutions. It is many of our experiences that Solaris x86 is installed on existing systems running the competitions' operating systems and closed applications. Once we transition these systems to Solaris and open solutions, Sun quickly proves to have the superior technology. Our customers are subsequently much more receptive to Sun hardware solutions as they outgrow their x86 systems."
Guys, it's the 21st century. Linux is more open still and every bit as good as Solaris on Intel. Don't like Linux's SVR4x style? Then use the more Solaris-like Berkeley operating systems instead. That's what the administrators who love Solaris but can't afford it I know are doing.
Another thing I find annoying about their arguments is that they want to keep Solaris on Intel second-rate by locking it down to the 32-bit architectures. There's already shipping commercial Linux from Red Hat and SuSE for such 64-bit architectures as Intel's Itanium and POWER. And, for that matter, alpha Solaris code was running on Itanium according to Sun back in October of 1999!
If Solaris-on-x86 was really serious about Solaris on Intel as an operating system instead of just a way to bait and switch way to get people to buy SPARC systems, they'd be arguing-as MacOS users have for years-that Sun could profit it from their operating system alone if they simply made it available on cheaper, commercial hardware.
All Solaris-on-x86 is really doing is trying to make a public debate of an internal Sun business decision; one, which, given that Sun continues to look at Java and SPARC for profits, makes perfect sense for Sun.
IBM never gave in to the pressure of its vocal OS/2 fans because there was no profit in it-OS/2 bottom line: because ISVs could develop for both OS/2 and 16-bit Windows using 16-bit Windows, they never had a reason to code just for OS/2 -- and Sun won't, in the long run, give in to the Solaris on Intel fans. Will Solaris go Open Source? No, that won't happen either. Again, where's Sun's pay-off? There isn't one.
Unix on Intel belongs to Linux, SCO's Unix offerings and the BSDs. Solaris on Intel's day is done and rehashing internal Sun business decisions in public won't change that.
For more on why Sun did what they did at the time see: Solaris on Intel Out? Does that mean Linux is In?
For more on why Sun did what they did at the time see: Solaris on Intel out? Does that mean Linux is in?
This article was originally published at Practical-tech.com. Steven J. Vaughan-Nichols is editor there.
Comments
on Cyber cynic: Solaris on Intel -- forget about it alreadyNote: Comments are owned by the poster. We are not responsible for their content.
Anyone can afford a SPARC system that can run Solaris 9, to test. An Ultra 1 just sold on ebay for less than $150 w/ monitor, and Ultra 5s go often for less than $500. I saw a beautiful Ultra 10 go for $600 w/ 19" trinitron monitor at a computer show. These are no longer top of the line, but they are more than enough to learn Solaris 8 or 9 on, and they arent any more expensive than buying Intel equipment.
Furthermore, you can buy new SunBlade 100s w/ Solaris 9 for for about $1000-1500 from Sun, refurbished for less.
Look around just a little bit, and you will find a ton of SPARC hardware capable of running Solaris 9 well, and most of it is not much more expensive than x86, if at all!
In fact, that's what I've done.... nice and easy, not just sysadmin and development, but writing shell scripts is something sysadmins do a lot of, and shell scripts very easily fall over if you developed them on one Unix (or, worse yet, GNU/Linux) and just expect them to run on another.
Not saying GNU/Linux is necessarily worse, but you cannot assume that everything works the same - GNU tend to add a *lot* of switches to common utilities, and slightly change the functionality of others. This is generally an improvement, but does not improve the lot of the muli-platform shell script developer!
--Sun SparcStation20 (32-bit SPARC) running NetBSD, $150 at a computer show w/ 19" trinitron;
--Sun Ultra 1 + creator 3d (64-bit SPARC) running Solaris 9 (a little slow, but usable), $180 ebay, shares the ss20's monitor;
--SGI Indy (MIPS) running Irix 6.5, $140 at show w/ 17" trinitron (an amazing little machine considering its almost 10 years old);
--Alpha 533MHz (64-bit) running Linux, $100!!! at a computer show;
--dual P2 xeon 450, put together over several years for a few hundred dollars, running FreeBSD;
--and finally, a fully tricked out dual athlon mp 1600 rig, geForce3, scsi raid 5 array, 2 GB of DDR SDRAM... (okay, i admit this last one wasnt done in a budget-consious way, it was a grad-gift I gave myself, but hey, $1700 is damn good for that kind of performance), running Linux;
--The alpha and 2 x86's share a keyboard monitor and mouse via a digital KVM setup, $100 but that was worth it to not have 3 different sets of those.
I would consider all of these to be worthwhile investments that have already paid off. I just got a generous job as a Unix sysadmin (Solaris and HP-UX) (without a Sun or HP certification) and was later told that this was because of my familiarity with the quirks of non-Intel hardware, experience I would not have had without my good old ss20 and indy (and recently the ultra 1).
Solaris isn't really a good choice now anyway
Posted by: Anonymous Coward on September 18, 2002 08:19 PMNow let's talk about Solaris in general.
To get an LVM as everyone thinks of it or a real JFS you must buy products from Veritas, Sun has yet to include these bare necessities (I feel like singing) when every other major UNIX based O/S, AIX, and HP-UX has included it for years. Now every modern Linux distro (even RedHat) has several choices for JFSs and at least one kind of LVM out of the box. Okay you can encapsulate the O/S with the LVM, but the drawback is IF you want to perform an O/S upgrade you have to un-encapsulate the O/S before you begin, which means if you ever took advantage of the LVM for the O/S you now have a mess on your hands.
To add disks onto an EMC cabinet, if one follows the recommended configuration from EMC, you end up adding disks to the sd.conf and taking an outage where the other O/Ss don't need to. It seems that the recommended configurations for Sun on a SAN is hardcoding everything, binding WWN to HBAs, disks,<nobr> <wbr></nobr>...etc.
According to Suns recommended procedure for applying kernel patches, one must be in single user mode for an extended outage. (Read their procedures or talk to the Sun support folks sometime.)
Oracle has announced that they will be running their business on Linux. Larry Ellison was pretty blunt about the future of Sun. What do you want to bet Oracle's base port for their RDBMS on UNIX is moving to Linux.
Sun has recognized that they're behind, CDE's days are numbered, so Gnome (or perhaps KDE) will be, if it hasn't already, the UNIX desktop that Solaris will include. SPARC is still fairly expensive, and when you start including the price of Veritas's products they're more expensive than HP!
Solaris's only saving grace is that they reboot faster than our other UNIX boxes!<nobr> <wbr></nobr>;)
Why even bother with Sun as a UNIX vendor? They're playing catch up with GNU, Linux, and the BSDs.
My current recommendations for our datacenter of over 225 UNIX boxes are HP-UX/AIX for Oracle (depending upon what we need) and Linux for the application servers where possible.
Re:Solaris isn't really a good choice now anyway
Posted by: Steven J. Vaughan-Nichols on September 18, 2002 09:28 PMThe slowest OS I ever saw was Xenix on a 4.77MHz 8086. Now That was Slow!
Steven
Editor, Practical Technology
http://www.practical-tech.com
Re:Solaris isn't really a good choice now anyway
Posted by: Mandrake Magician on September 24, 2002 02:46 PMWhere I got my only formal Unix training, they give the students a diskette copy of Coherent, talk about Unix and run Linux on the 486-braindonor server.
I'm not certain, but I suspect the Unix program was put together by a MSFT certified reseller.
Your Solaris experience is obviously years out of date, and you are assuming that nothing has changed since then.
LVM: Now part of Solaris 9, DiskSuite has been free for years, and at least as good as Veritas. Solaris 7 onwards have a journalling filesystem built-in, too. Just add "logging" to the vfstab options.
You can also upgrade a DiskSuite / LVM-managed OS disk to Solaris 9 without having to think about it - when you reboot into Solaris 9, all your mirrors, RAID, etc, are there as before, even the OS disk(s).
If you knew anything about SANs, you'd know about WWNs and why they are used - look into MPxIO (another reason you don't need Veritas, whose version is DMP)
Of course you should be in single-user mode for a kernel upgrade. For most OSes, you must shut down the machine.
If single-user mode is not possible, the machine should be as quiescent as possible. That's just common sense.
CDE is not just Solaris, it's HP, IBM et al. The COMMON Desktop Environment. Sun are moving with GNOME (well, I guess you can say it's more recent, at least!). GNOME1.4 is bundled with Solaris 8 as an unsupported beta, 2.0 is supported with Solaris 9.
Re:Solaris isn't really a good choice now anyway
Posted by: Anonymous Coward on September 19, 2002 11:44 PMObviously based upon your comments you are fairly well focused on Solaris, perhaps not on much else.
UFS was designed to be journaled from the ground up?<nobr> <wbr></nobr>;) No, the journaling was "bolted" on top of UFS, much like ext3, while it is nice to say look we can do journaling, it is another to say it was designed from the ground up to be journaled, versus bolt on journaling. I wouldn't trust ext3 for mission critical data, but I would trust things like reiser, sistina's GFS, JFS, XFS, and vxfs. I believe my comments were in reguards to a real JFS.
At least with Solaris 8 Disksuite was a far cry from the OSF based LVMs that HP-UX, AIX, True64, and Linux all use. It reminds me of "mirror-disk" (md) on Linux. It appears that perhaps I didn't give it enough credit, it looks like it can do a lot of stuff with it, but it still blows comparatively, it appears to be a lot more complex than it needs to be. Can you directly install the O/S to an LVM managed set of objects, or do you have to convert after the fact?
As far as SANs go, I know why WWNs exist, obviously this is an area where you really don't have exposure to the other O/Ss. I'm not talking about zoning, or switches or anything else. I shouldn't have to tell the host O/S jack about WWNs for anything period. The only O/S so far that I have had to do this with is Solaris, well in the recommended configuration from EMC anyhow.
As far as kernel patches or upgrades are concerned the only O/S that I have yet to use that wants to be in single user mode to apply (as opposed to pick up) the upgrade (or patch) is Solaris, on all the other O/Ss I can be in multi-user mode and bounce the O/S when I want to pick up the changes. HP-UX is a bit on a pain to deal with in this area, but it is still better than Sun's recommended approach. And no it isn't common sense that one cannot change a set of kernel files on disk that are in memory.
You are right that CDE is being tossed aside by all the vendors for a an opensource desktop. But why stop there? The O/Ss in the opensource world have caught up and surpassed many of the "legacy" UNIXs<nobr> <wbr></nobr>;) for general purpose stuff.
Based upon your comments you must either work for Sun, a reseller, or you are in a Sun only shop.
Re:Solaris isn't really a good choice now anyway
Posted by: Anonymous Coward on September 20, 2002 08:32 AMRe:Solaris isn't really a good choice now anyway
Posted by: Anonymous Coward on September 22, 2002 02:24 PMThough I will stand up to the robustness of the likes of Solaris and *BSD, Linux has many merits... and many of them are reimplementations of ideas that are deeply rooted in the Unix culture in a totally new way. (read: innovation of ideas)
Reiser was a mistake to include into a production kernel, IMHO. It was complaining mostly that got it in there, and somehow just being in there makes people use it. If your running anything that matters to you, my advice would be to go with another filesystem that has some nads to it.
By the way, it doesn't matter where an original idea came from, considering both XFS and JFS are being re-written from the ground up by their respective distributors. (read: SGI)
Re:Solaris isn't really a good choice now anyway
Posted by: Anonymous Coward on September 22, 2002 02:13 PMI wouldn't trust ext3 for mission critical data, but I would trust things like reiser, sistina's GFS, JFS, XFS, and vxfs.
-- end quote --
Any person who, in the same breath, says they wouldn't trust mission critical data with ext3, but *would* trust it with reiserfs, really needs to test it out a little more. I have ran reiser on 3 occasions, fully trusting it with my critical data. On all occasions it has failed me... corrupting itself to the point that even reiserfsck could not repair it. Every instance that this has happened, I took the same data (luckily, I do run backups) put it on XFS and ext3, and everything ran smooth as silk for much longer than reiser did. (and still going)
ext3 might not be a speed demon, however it definately does it's job very well. Adding data=journal to the options in the fstab makes it very robust.
Reiser looks good on paper, but in the real world it definately does not shine for production usage.
Regards<nobr> <wbr></nobr>;-)
victor
"All men dream: but not equally. Those who dream by night in the dusty recesses of their minds wake in the day to find that it was vanity: but the dreamers of the day are dangerous men, for they may act their dream with open eyes to make it possible."
— T.E. Lawrence
Seven Pillars of Wisdom
There has to be one way of sampling Solaris
Posted by: Anonymous Coward on September 19, 2002 05:13 AMThis is the #1 reason not to use a close-source OS
Posted by: Anonymous Coward on September 19, 2002 11:29 PMI should also point out that the latest version of Solaris (currently 9) is backward compatible with the Sun SParc CPUs of the late 1980s. Sun has supported all their hardware and this is one of their selling points. So for each new software evolution or revolution sun makes sure its customers are kept in the loop.
Solaris for Sparc and Intel with Code that is 95 % identical to each other with respect to the hardware difference. I said code that is the same NOT binary ok. So the Sparc and Intel version of Solaris has ALWAYS been based on System V R4 to be specific.
Linux is actually more BSD than it is anything else. Based on the Mimix OS. Look at the following diagram to help your memory some on that http://www.levenez.com/unix/history.html#07
Now I have been using Solaris for long time even Sparc's. I today wouldn't be without my trusty Solaris x86 laptop.
Linux still lacks a lot of what Solaris can do and many programmers who do a lot of porting between OS's will attest to that.
Solaris x86 is far more stable than what Linux will ever be and mostly definitely put Windows to Shame.
At least with Sun I can count on getting Solaris regardless if it is on Sparc or Intel. With Linux there are so many different variants and everyone of them will drift off to their own tastes and branding (Red Hat, SuSE, Debian, Slackware, Mandrake) may I say more.. They are so many different brand and versions, It is any different
with current unix vendors who differ in their OS branding of Unix HP-UX, IRIX, AIX just to name a few and now Sun wants to add their flavor..IBM and HP has their flavor of Linux. Then next they are all not in sync.
At least before 1/8/2002 announcement with Sun, Solaris x86 was in SYNC with the Sparc version
and nearly all features of Sparc were put into
Solaris x86 as well. There are exceptions and the
Solaris x86 community recongizes that and have been working with Sun to rectify that.
No, I'm not one of the secret six but I'm definitely one of the longest standing members of that list.
You also have to realize by cutting Solaris x86 back like that affected many Sun Customers who are currently depending HEAVILY on Solaris x86 in their core environment.
Yes Solaris x86 had it's slow start with 2.1 but has grown really well while in sync with current realease of Solaris ever since 2.4. with the except of this last release 9.
Linux still has a long way to go in the SMP world yet. Ever stop to wonder why NCR went with Solaris? SMP
Another thing, how offend can you look up a man page on any commands that are strictly to Solaris
on Linux. The Answer is YOU CAN'T. But I can guarantee that you'll find them all there on Solaris x86.
Phillip Bruce
linux is not solaris, nor a suitable replacement
Posted by: Anonymous Coward on September 20, 2002 05:49 AMSpoken like a linux-only admin.
I work with sun's high-end boxes, as well as low-end boxes. E10ks, to netra T1s. I also work with Solaris x86. Admining a linux box is fairly different from admining a solaris box. On the other hand, admining a low-end solaris sparc box, is virtually identical to admining a solaris x86 box.
I use linux where appropriate. At my workplace, there is no gain to using linux there. There IS gain from using Solaris x86 there; Less hassle for me and my co-workers.
Solaris users are badly misinformed, and there are
a number of serious factual errors in the article
itself.
Rather than belabor the issues, it really doesn't
matter what the writer thinks, since there is an
obvious market for the robust and reliable Solaris
X86 OS. Sun has multiple-millions of existing
Solaris X86 users, and is trying to force them on
to their Sparc OS, to shore up it's failing sales
on Sparc. That's pretty clear to anyone who really
works with Sun systems as a developer.
They didn't think that the existing millions of
users (many of whom paid hundreds of dollars for
legal licenses for the OS, and were willing to, to
obtain a good product), would complain about the
OS. This is because even Sun's senior management
doesn't pay much attention to the X86 product, and
doesn't know how much it's changed since it's
inception in 1993 or so (version 2.4).
But there are many people who really understand
the use of Solaris on Intel, and are willing to
fight for it's continued availability. For those
who don't, who cares - go use something else. To
try to change the minds of people who are happily
using Solaris X86 seems to indicate that the
author has an ulterior motive for his article. It
is certainly not hurting him that Solaris exists.
So what's his point? Doesn't want to hear the
arguments? Fine, then don't listen.
For those of us who badly want to see Solaris X86
continue to be available, we will continue to work
to show Sun what they are missing. We certainly
don't need Sun to waste money creating and
supporting _another_ LINIX version. What's the
point of that? There are lots of LINUX versions
to choose from. Doesn't the author want Solaris
users to have a choice too? Hmm, wonder who he is
buttering up with this type of article?
If he really wants to increase his readership, he
would do better to make sure of his facts first.
Then he might be worth another read. As it stands,
he made so many errors of fact in the article that
it's not worth reading at all, since it doesn't
give a unskewed vision of the situation.
If you really want to know what the issues are,
read the open letter to Scott McNealy in the SJM.
--an Solaris developer
Solaris 8 on x86
Posted by: cykes on September 18, 2002 07:44 PMAlso Solaris users many who can not afford sparc systems as personal systems are more likely to encourage their employers to use Sparcs just as MCPs and MCSEs are always trying to change "working" (emphasis on working) Unix infrastructures to Windows environments.
In the end Sun can't lose. With Solaris 9 out, Sun can easily donate Solaris 8 x86 to colleges and schools etc. similar to MS and Redhat's methods for gaining mindshare in the up and comming future developers.
#