Slashdot Mirror
Sun Considers Switching Cobalt to Solaris
cyber-vandal writes "According to this story on ZD in the UK, Sun is considering switching from Linux to Solaris, even though Solaris doesn't actually run on the MIPS architecture." "Ok guys, we bought this company that seems to be doing ok and we want to drive it into the ground as fast as possible. Suggestions?"
Solaris on the Cobalt RaQ and Qubes might not be a bad idea in many ways, from both Sun's and Cobalts point of view. Also, it could benefit the end user as well. Consider the following, just for a moment.
1) Sun needs a foothold the low-end market. A conspiracy theorist could see that Sun is just trying to expand their empire, and with some just cause. Though Sun's competition has been makers of other higher-end systems (DEC, IBM, etc), all these companies have been feeling a loss to x86-based servers running Windows or Linux. In some ways, this has put a very strong line between who rules the high-end and who rules the low-end. However, when it comes to the bottom line, Sun could be making mroe because it is easier to sell more cheaper boxes with a cheaper OS. So, Sun does a couple of things: Make Solaris 8 free (sans media cost) for under eight processors and increased support for the x86 platform as well as addming many popular GNU tools found in Linux, FreeBSD, etc (an ironic fact if you looks at the history of Solaris or SunOS). Now, you say Solaris has x86 support. Yes, but it leaves something to be desired. However, something Sun knows how to do well is write to a very specific set of hardware. Sun writes Solaris for UltraSPARC based servers. There is only very variance in hardware, unlike the x86 world, where it is a hardware goulash. So, Sun takes Cobalt, which is very inexpensive (if you have ever looked inside a RaQ, you'll hate yourself for not coming up with so simple five years ago) *and* only uses a certain hardware. It would be easier for Sun to optimize for the RaQ than to try to make every x86 user happy. They end up with a more stable x86 offering and don't have to change a thing to the already cheap hardware. They are again competitive with the real leaders in the Server market. Face it, Sun can only sell so many mammoth Sun boxes nowdays. Talk to any Sun sales rep. They feel the pinch badly. So in the end, Sun gets with the times in a buisness sense. Produce a inexpensive, stable server with the Sun name, and not have to re-invent the wheel. Though it is too early to see, Sun could be making a step in the right direction an it could also solve a problem the Cobalt is having, stability. As for OS supremacy, Sun makes no real profit in that Microsoft since on their OS. Sun makes hardware. That is their bread and butter, not this stagnantly evolving mish-mash of BSD and System V. Sun changes Solaris to work better on their hardware. They are not big "feature" mongers.
2) Cobalt needs Sun. Now, before you begin to flame me for possibly hinting that Linux uis unstable, that is not what I will be talking about here. Cobalt made a mistake early on, and it wasn't choosing Linux, it was not making a strong push into a hardware platform. Cobalt started on MIPS and went through great pains to make their special version of RedHat 5.2 and assorted packages run on the MIPS processor as well as their highly-flaunted web-based administration software. By RaQ2, they got it somewhat stable and really started selling. However, their were issues with their administration interface and theiur lack of 3rd party support. One was do to a software development platform, the other, hardware.
Yes, the Cobalt "special sauce" is very perl based, however, their is a nice helpinf of servlets about. Though IBM has made great strides with Java on Linux, they weren't around at the time. The Perl/CGI and Java combo on Linux, from by point of view as a developer, is a match made in hell. I wouldn't do it on any other OS. Also, the JVM support wasn't that great on MIPS either. Overall, scalibility suffered on these boxes. Also, if you ever dared to go beyond their static administration environment, you broke that special sauce. It is not a picnic to fix. Combine this with the fact that many applications (RealServer, Oracle 8i, iPlanet/Netscape servers, Sybase, etc) that run fine on x86 Linux, could not run at all on the RaQs due to MIPS. Cobalt made the change to x86. The migration was an absolut nightmare. They drop all MIPS support on the RaQ3 so they can get better Java support and support more 3rd party apps. Well, they never coudl get a migration package to work between the older RaQs and the newer ones. Cobalt didn't write Linux, of course, they knew how to make cheaop hardware and a very *large* and convoluted web app. Cobalt planned an easier migration to x86, however, never considered all they needed to do to revert the "altered RedHat" back to x86 and make the according changeds to their app. This seems like it would be easy, however, it seems that Cobalt made things harder. I don't believe that they stuck with 5.2 as their base for trhe RaQ3, and it the structure changes broke all their platform-indepedant code they chose not to rewrite very well. Cobalt sales go down since they don't support this mess very well, as they made very little changes to fix this. Cobalt made bad desicions in changing to much of their original formula. Their product suffered. Along come Sun, with an OS that rarely changes general format. Cobalt can do what they do best: make cheap hardware and write a very user-friendly interface to a OS that supports all those lovely 3rd party apps the customers want. Cobalt's formula doesn't take well to massive hardware or software changes. They made that boat. Sun could possible save it from sinking. Was it the best boat? Maybe not, however, Cobalt really wanted the whole thing to be cheap and have the customer touch little of the OS as possible. Cobalt *could* have made their product work just fine on x86 and Linux and made it very stable. They chose not to make that effort. Now, the Java support will be top of the line. Sheesh.
3) Lastly, the customers. These are they people that actually matter in this whole thing. The people who by this to do what they want without having to be geeks. People who want their little mom and pop internet shops or web hosters. They pay for these Cobalt boxes and don't want to worry about their nice user interface not working with the OS or worrying that more hardware changes means they can't upgrade as needed due to bad implementation. They change could benefit the customers. The majority of Cobalt customers don't mess with the OS, since it breaks their pretty interface. This important, since these people don't want to be geeks, they just want get a server online cheaply with the smallest learning curve and have it work right. This could solve that. This is not meant to be a geek's server. This is your mom or dad's server. When Cobalt can't make these servers upgrade smootly, mom and dad cease buying more servers. This *could* be remedied now.
Closing: Sun keeps the RaQ on the x86 path with a static OS. Sun gets their foot in the low-end door. The customers hopefully get what was advertised in the beginning: a low-maintainence, cheap, stable web server that is easy to run and that doesn't require hundreds of hours of pouring over FAQs and arcane tomes to administrate. The catch: Sun. Sun will be new to this market. They could take Cobalt's implementation and make it worse by doing the same thing that Cobalt did: change without forethought. This would include price increase. That would be the worse thing Sun could do, as it would show that they just don't get it. Sun has the ability to make Cobalt's implementation work, as long as the customers get what they pay for in the end. Unfortunatly for Linux fans, it will be without Tux. Could of this been avoided? Yes, easily. Sun may of never came into the picture if that were the case.
Here are some background articles on the purchase:
Upside Today
Red Herring
Morningstar
Reuters.
Work for Change & GET PAID!
Cobalt was dropping MIPS support long before this, going to pure Intel. Solaris already has Intel support, and is actively supporting the Intel versions.
Solaris on Intel is 99.9% the same source code as Solaris on Sparc. Only the kernel and 64-bit support differs.
I do not deploy Linux. Ever.
There is a strong mindshare among Solaris admins that Solaris is a real UNIX, and that Linux is not a "real" UNIX. Sun makes a good deal of money with this perception. Sun is not foolish--they know that if management types realize that low-cost Linux servers can do as good as a job as Solaris for a large number of intranet and internet tasks[1], they will go for a low-cost Linux investment instead of a high-cost Solaris investment.
As it stands right now, Linux is a far greater threat to Sun than it is to Microsoft, since Linux works well as a UNIX server, but yet is not as functional as Windows on the desktop[2].
- Sam
[1] There are things, such as NFS and support, where Sun/Solaris is better, of course.
[2] We're getting there. I think Kde and Gnome 2.0 coupld very well put us there.
The secret to enjoying Slashdot is to realize that it should not be taken too seriously.
- more stable
- as fast or faster than
- as supported as
Linux.
This is a case of corporate pride leading to poor decisions.
Sounds more like a case of "The previous article is a case of fanaticism clouding judgement".
As others have pointed out, cobalt was moving to x86 anyway. Doing a comparison on that basis,
1. Solaris is certainly not going to be LESS stable than linux.
2. There's probably not going to be more than a 10% performance difference either way.
3. How could solaris be "[less] supported [than]" linux? By SUN? That's just silly.
Just a few months ago, I worked for an ISP that transitioned over 1000 commercial web hosting customers from a DEC box and a sun E450 to 3 Cobalt RAQ's.
Within DAYS of making the transition (painful to be sure), we were already hacking apart the way the RAQ's handles some of the CGI, and apache aspects of site hosting. We were also making custom scripts for system admin tasks that we needed to change some from the way the RAQ's web admin interface did things.
While this could be done in solaris as well, it would have taken a little more work, and of course someone more knowledgable with solaris. We had that manpower, but not all people do. Plus, if you are forced to switch your hardware to Solaris for continued support, you then have to fix any and all custom work you did, with most likely LIVE customers, and in a hurry.
As they are, the Cobalt boxes I have worked on are a BREEZE to configure and maintain, I would like to ask this VP, WTF is the point in changing it?
Shameless Plug! If you like trance, tell me what you think!!
Ok, ok...so the title might be a bit trollish/flamebait...
<disclaimer>I used to work for a Cobalt Authorized Reseller, and I didn't like their gear </disclaimer>
If you've ever tried to use a cobalt machine to anything besides what it is designed for, you're in for a treat. PERL is usually broken (have to manually fix files that are standard in most other PERL implementations). There are a number of bugs in their web interface, and their implementations of most services are non-standard or just funky enough to make it pretty annoying.
The cobalt mentality is also pretty closed in general and support is lacking in response time.
These boxes aren't built for hackers, their built to be canned solutions for Linux neophytes and web-farm admins that don't want to get into the nitty-gritty. Switching to solaris would be a net change of 0 in my book, as most people using cobalt equipment don't care anyway. If you don't want a toaster-like solution, don't buy Cobalt Networks gear...it's underpowered and overpriced.
This comment is wild and rambling. I hope someone else says something similar, but more coherantly. I apologize.
-fp
Ok I guess we've all decided to give up the 'right tool for the right job' idea and ram linux down the throat of every system and every problem:
Ok guys, we bought this company that seems to be doing ok and we want to drive it into the ground as fast as possible. Suggestions?
So if they don't use Linux, its bound to fail is that what you're trying to say? Solaris is a great operating system and for some things its better, yes I said better, than Linux. Is Linux better than Solaris at other things, yes it is. But I've yet to read a post in the comments that has any insightful revelations as to why Linux is better than Solaris on the Cobalt machines. Just a lot of Linux hype. I'll back up the fact that maybe they should go to Solaris with some facts:
1. Sun will stand behind and support their OS with one of the worlds largest software engineering force. If a customer had a problem with the Linux OS what would their options have been? Post to a newsgroup? 'Ask Slashdot'?
2. Solaris has tools that make developing operating system level software for a new platform that Linux can only dream of right now...ie the Solaris Kernel level debugger.
3. Solaris is proven. While this is some fact and some opinion the truth remains Solaris/SunOS were running stable and secure servers when 99.995% of the world had yet to hear of Linus Torvalds.
It just really disheartens me when I hear everyone screaming how monopolistic Microsoft is and how blindly people follow them and then I read comments as I did here. For those of you who made comments like the one I quoted above without fact to back it up, you're no better than the M$ fools you make fun of on these very pages. Blind advocacy does nothing to support the OS you're so fond of. For those of you who did use reason and facts to back up your comment/opinion on the subject, I apologize for the rant.
And for the record, I do use Linux at home and encourage Solaris usage for the servers at work.
Clearly you have never used SolarisX86.
"2. There's probably not going to be more than a 10% performance difference either way."
See answer above.
"3. How could solaris be "[less] supported [than]" linux? By SUN? That's just silly"
Spoken like someone who has never had to wait on hold for Sun technical support (who they were paying through the nose), or eat dirt over a bug Sun has labeled "SORRYWEDONTFEELLIKEIT" in their bugbase...
Dont just naysay Linux. Especially if you've never even tried the competition or made a serious comparison. On MIPS, Solaris doesn't exist, and believe me, is very unlikely TO ever exist. That's probably a good thing. On X86, Solaris is a JOKE - with terrible app support and a really flakey, bloated kernel. Let's not even talk about security out of the box, let alone the awful way they've configured it or how INCREDIBLY expensive eveything is... let alone your precious "corporate support"... and, last but not least, Linux smokes Solaris on sparc hardware.
Solaris is the Windows of the Unix world. I have nothing against Sun, I love their hardware and I am a huge Java fan. But their Unix... blows.
Put that in your crack pipe and smoke it.
We're on the road to Tycho.
We went with Cobalt because they offered incredible integration and preloading services crafted to our specifications. If the OS switches to Solaris, lots of our existing software won't run without convincing our suppliers to compile it for SolPliance OS or whatever they want to call it. If they even _let_ anyone do that.
Good thing we're investigating other suppliers of point-and-drool rackmount servers.
Yeah, but they're trying to unify them. That's what Windows Milennium is about, migrating users to a single OS that will come in editions that range from Professional (desktop workstation) to Data Center (high-end OLTP and DS). That's the plan, anyway.
IBM manages to drive OS/390, Linux, Windows, AIX, OS/400
The reality is that AIX and OS/400 divisions of IBM actually compete with each other, and that IBM are huge enough to have the resources (and the locked-in legacy customers/revenue) to hedge all their bets. Besides, if you need an S/390, then you really need it, and nothing else will do.
SGI manages to drive Irix, Windows, Linux
Must be some definition of "manages" that I'm not familiar with :0)
HP manages to drive HP/UX, Windows, Linux
OK, that's a fair point, but see IBM. HP are huge and have their fingers in many pies.
Compaq manages to drive Tru64, Linux, Windows, VMS (or is VMS dead yet?)
VMS is far from dead, and Compaq is still a roll-up of Compaq and DEC.
Why does Sun not have the energy to drive more than one operating system when its competition does
If they can unify on a single OS, then they potentially have a strategic edge that their competitors can't match. For example, SGI are busily trying to port Irix features to Linux, Compaq want to move VMS features into Tru64. If Sun can add features to all it's platforms simultaneously, that's reduced time to market, and higher quality right there.
Might help the boxes gain acceptance in shops that are mostly Solaris.
Here's how it is on a Raq. Each virtual site directory is owned by httpd, and owned by a group created for that site. httpd, admin (the webmaster basically, you don't use root to maintain web content), and any site users belong to that group. Works well in theory, but once you have more than 30 virtual sites on one raq (there's one here that has 56) then the admin account stops working, and you need an admin2 to belong to the higher numbered groups, then an admin3, and so on. This is because you can only belong to 32 groups in Linux -- and only 16 in BSD, and the same problem probably exists in Solaris. Any more groups after that are ignored. This is the pie in the face of those who believe groups are adequate to replace ACL's.
... none of those are here now, and til then I am kludging around the situation in various ways, none of them particularly elegant. I'm not even fond of POSIX ACL's, since you can't centrally edit their definition like you could on even PRIMOS. Imagine a user-defined group and you're close.
Solaris has ACL's which just don't have that problem. You can put an ACL on a dir, and create the same effect as the setgid bit with it as well, having all files and directories created under that dir also inherit it.
Yes, yes, ext3, xfs, jfs
Normally I'd jump at the possibility of solaris on a raq, but I really don't care much for the interface. Webmin is already far more functional, faster, and less cumbersome. And though cobalt is a little closer at having its interface integrate the tasks of managing virtual hosts and their DNS records, its interface is simply too slow to bother with. Raqs are nice for office workgroups, but they may as well just be another form factor for a qube, and just as bad a fit for a ISP.
I've finally had it: until slashdot gets article moderation, I am not coming back.
C'mon . . . no one saw this coming? A big company buys out a smaller company and then trys to incorporate it's products into the new domain. It's the way of the business world. Maybe it'll suck (read: Hotmail) or maybe it will actually work out. Solaris isn't really that bad if it's administered right.
--
Sun doesn't care about Linux. They never did. Nor open source software. They GPL'd Star Office because it wasn't worth paying developers to work on a product that can't compete in the Windows space. Or maybe to be a thorn in Gates' side..
Sun probably has a marketing dept. that insists on being able to say their server appliances are powered by the same 'high-powered' OS that the big iron is.
I have nothing against Solaris, there are a few features I actually like, but I don't see how a new Solaris port to MIPS could possibly be:
- more stable
- as fast or faster than
- as supported as
Linux.
This is a case of corporate pride leading to poor decisions.
Marc
Although everything before the raq3 is MIPS and had many problems with software compatability, my raq3, and the raq3i and raq4's are all x86's running linux. Fairly straight forward compatability. Don't think that solaris is such a hot idea, though....
My sentiments exactly. When did Linux become so much better than it's brother unix, Solaris, that we simply assume everything has to be Linux?
Oh, and a good suggestion to drive Cobalt into the ground? Forsake actual developement of new technology in order to support GPL'd versions of all often done old technologies. Like Unix.
-Ben
One of Sun's senior VPs (Andy Ingram) said:
Why does Sun not have the energy to drive more than one operating system when its competition does?
I'm a big Sun fan. From what I've seen only IBM (with AIX) can compete with Sun in the enterprise class open systems niche (though HP might have something with Superdome). And I don't know that the move to put Solaris on the Cobalt line is necessarily a bad thing. However, I don't understand the suggested reasoning put forth by Sun at all.
have a day,
-l
Find the manager for eToys and hire him.
--
From the article:
"Cobalt's [devices] are true appliances," Ingram said. "You do not care about the operating system. If we were to convert these over to Solaris, the end-user wouldn't care. We don't have the energy to drive two operating systems."
Right. One of the reasons that I bought into Cobalt's products is was that I COULD buy them and set them up without a fuss, but the option was there to modify the OS and software even though it wasn't supported.
I fear that if they port Solaris to the product, I won't have that kind of access anymore.
As for porting it to MIPS, that's only for the Qube series. The RAQ line of servers runs AMD processors. Solaris x/86 will fit, I think (though I've never run it on anything other than Intel chips).
"All those tubes and wires and careful notes!"
"All those tubes and wires and careful notes!"
This move is obviously part of the Sun overall strategy. For 3 main reasons:
1. They make Solaris. It makes sense for them to try and run everything of theirs from this platform. Their programmers know the OS inside out. This move is about as surprising as hearing that Microsoft is trying to move Hotmail over to Windows 2000 from BSD. So we'll see if Sun has any better luck than Microsoft.
2. Announcing this OS move is also obviously a publicity stunt designed to try and put forth their own Solaris as a superior OS to Linux. Remember, like Oracle and Microsoft, Sun has a major superiority complex.
3. In porting Solaris to another platform they are improving the overall portability of the OS, making it a more attractive OS.