“The news is a credible, coherent Linux desktop from a major player. Despite their big bets on Linux, neither HP or IBM has ventured into this territory. The combination of the desktop, the pricing model, and Sun’s promise to gut the middleware market is bad news for Microsoft, IBM, and anyone else that makes money from the complexity of enterprise systems. If this strategy gains even a little traction, it will start a pricing downdraft through the whole industry.” Read the article at NewsForge.
Sun’s Software Revolution
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Sun’s Schwartz Speaks Out on Linux, SCO
September 19, 2003
“Also, let me really clear about our Linux strategy. We don’t have one. We don’t at all. We do not believe that Linux plays a role on the server. Period. If you want to buy it, we will sell it to you,”
http://www.eweek.com/article2/0,4149,1274614,00.asp
Well, I have to admit, given that statement I wouldn’t really give my money to Sun for whatever Linux products they sell given that to them it’s the unloved stepchild. I wouldn’t believe that they are really comitted to it.
Why is it called a “Java” desktop? Is it coded in Java? If so, how much of it is Java code?
http://www.osnews.com/comment.php?news_id=4591#145222
Please don’t recycle the same conversations.
You might want to read this.
http://www.eweek.com/article2/0,4149,1300161,00.asp
Schwartz Seeks to Clarify Sun’s Linux Strategy
raptor: *shrug* there isn’t enough trust there after the first interview to validate me putting any time and effort towards their linux products while other companies don’t change their public opinion on Linux once a week like Sun does.
They putting their focus on Linux, however they don’t believe you should be buying Linux. They plan on taking 10% marketshare in the first year alone, but with such a negative attiude twoard the product your selling what would make anyone want to buy it? With statements they make, I don’t think companies are going to put their trust in Sun because Sun doesn’t trust their own products.
What Sun is creating is a system. We all may critical of Sun and their intentions with Linux, but they have tremendous respect in the IT industry. They add validity to Linux as a desktop to big corporations and governments. Think about the impact of US Social Security call centers changing to Linux desktops (we’re talking thousands of users). Or maybe the US Visa office of the state department getting rid of annoying windows worms and viruses for good.
We may all wish for companies to use Debian or Gentoo but the truth is, major IT groups want a big name company with support to role out products.
In the long run, I agree with the article that it should drive pricing down and simplify licensing for all platforms.
Cheers,
Sweetsdream
Sun’s Linux statements do not inspire confidence. If you are going to hawk a product, the least you could do is be enthusiatic about it. But Sun doesn’t just sound tentative, they almost sound hostile to Linux. Which is still allright, but I really don’t see any revolution in this new package they are offering. I don’t see anything compelling about it. Nor do I see any amazing strategy. All I see is Sun buying a lot of press time. Until they figure out what they are going to do about the maturing Intel hardware, Sun will continue to be the aging man on the block.
Eugenia wrote:
“http://www.osnews.com/comment.php?news_id=4591#145222
Please don’t recycle the same conversations.”
The linked conversation does not answer to the question. Does anyone have a 2 or 3 sentence answer to the “Java?” question?
Regards,
Mark
May I add that Sun has been paying license fees to SCO? I don’t know, at the moment I’d not like to sponsor SCO in any way, even not by paying money to companies they receive money from.
Sun appears to be trying to unify their branding under the JAVA name in much the same way as Microsoft did with .NET. There are no code changes to the components in terms of them being turned into java coded applications.
Can we please stop with the Sun is in bed with SCO nonsense? Why should paying for unixware driver licences have anything to do with Linux?
If you read the Article it is clear sun wants to revamp it’s Solaris x86 to be competitive with linux and to do that they need hardware support and since Unixware probably has a bunch of x86 drivers using a very similar driver model to solaris it would make sense to buy them.
What would Sun do make clean room implementations of OpenSource drivers and get into all sorts of litigation by the FSF for GPL violations and not to take into account hardware vendors suing for IP violations? How much would that cost? probably a lot less than the licensing them from SCO.
Think about it.
Do you know how much a comercial linux license costs from any of the major distributions? About $10,000 per CPU this is from an infoworld article. Infoworld is against SCO and very pro linux.
For the enterprise market linux is not free.
They don’t need to be in bed with them. They create revenue for SCO, that is enough for me to personally shun away from doing business with them.
10.000$? Whoever pays that was ripped off. We pay far less. Well, I guess there need to be stupid people out there you can rip off like that *shrug*
It’s because people are paranoid and ignorant, just like in a lot of KDE threads peopel say “Watch out, the Canopy Group, same group that owns SCO has a 5% stake in Trolltech, boycott KDE and Qt”. In reality, the Canopy group has stakes in dozens of companies and most of Trolltech is owned by Trolltech, this is just FUD.
Hearing these kind of things reminds me how stupid people are, it sounds just as pathetic as people saying “Boycott Mandrake because their French, and the French happened to disagree with America on the war, it doesen’t matter if they had any valid arguements, they disagreed.” This again is stupid, what influence did Mandrake have on their decision.
Please people let’s stop this nonsense.
SCO ledgers show that two companies paid them money — Sun and Microsoft, both to the tune of several million dollars. This is the *only* money they made in the last quarter. I really do not believe all of this is as simple as wanting UNIXware drivers. Sun doesn’t need to get drivers from UNIXware. Solaris is different enough from UNIXware that it wouldn’t be a simple port. And the x86 *BSDs have enough of a driver base that, if they wanted to, Sun could just use those.
People use more software than just OpenOffice, Mozilla and Evolution. They need custom apps. I think Sun wants to offer this desktop as the ideal platform for using inhouse-developed Java apps instead of inhouse-developed visualbasic apps or delphi or (anything) apps. I’m sure they’ll provide solutions for making easy deployment of such apps,and they’ll take care about integration of these apps with the desktop. I think that’s the “JAVA” part of the Java desktop system.
Sun is the only CONSISTENT and RELIABLE company in the whole IT world. Not the whore IBM, although I admit they have a lot of products which, in my opinion, are the best. But internal politics and living only today (they have to please shareholders) make this tremendous potential go to waste way too often. Not the HP — see above, plus junk for the masses. Yes, Sun charges money. They have to eat, too. And invest into future, by the way. I want to see you, children, grow up a bit, get off parents’ neck, and go to earn some money. Better yet, found a company. What a refreshing experince! Some parts of the world went through the stage of “our aim is communism”. They are going back to normal — most of them, and painfully at the same time! Just grow up! Stop being demagogic!
>>Can we please stop with the Sun is in bed with SCO nonsense? Why should paying for unixware driver licences have anything to do with Linux? <<
I don’t think it is “nonesense.”
Sunw was never interested in throwing millions of dollars at scox until scox started saying that: “all versions of linux are illegal – except sun’s version.” Sunw started the huge payoffs to scox at the same time as microsoft. Also, at the same time, sunw started getting boat loads of warrants from scox. McNealy has parroted many times since this began. Also, since this began, McNealy has been very fond of saying, essentially “buy from sunw, or scox will sue you.”
Put it all together with the fact that linux competes with solaris, and sunw is scared to death of linux.
Yarrow (Canopy’s CEO) sits on Trolltech’s board of directors. Trolltech own 1.5% of scox, and trolltech refuses to sell that. Since trolltech is a private company, trolltech could tell canopy that trolltech doesn’t want canopy invested in trolltech – but trolltech won’t do that.
It is somewhat surprising to me that people are such fans of some companies that they refuse to see the obvious.
>>If you read the Article it is clear sun wants to revamp it’s Solaris x86 to be competitive with linux<<
Why? I thought sun was behind linux for their x86 strategy? Can we count on sun to fully support linux when it directly competes with sun’s own product?
Linux woke Sun up about its semi-dormant x86 Solaris, which was mainly lacking in drivers for current x86 hardware (IMNSVHO, Linux’s greatest strength).
So – get some drivers “on the cheap” from SCO – if you read the “clarification”, the warrants were “just part of the deal”. Believe it if you want, or not; that’s up to you – I don’t know, and neither do you.
At the same time, if customers keep asking for Linux, it makes sense to sell it to them – Sun have been doing this with SunLinux, which didn’t really work because “is it a desktop? is it replacing Solaris? etc?”, so they clarify the position with MadHatter .
MadHatter is not just Linux, it’s an OE on Linux, Solaris, whatever. So far it’s Linux because that’s the easiest to play with. It’s not for servers, it’s not for power-users. It’s about getting rid of this “you have to spend $1k+ per desktop” bullsh*t which MS want us to believe. If a company has 5000 employees, of which 3500 use PCs, how many of that 3500 need more than email, web, and WP capabilities? I’d guess about 20%; the other 2800 can use MadHatter.
The stong remains the same; for a server, you need a real OS. That’s why we don’t play with MS crap (Thanks Eugina for that Win2003 article – things like “supports gigabit ethernet” were a real hoot!) – the server needs not just GE, but 8+-way SMP, real SAN and volume management support, which Linux just can’t do (yet?). For a 50-volume, 20TB 24-way database server, there’s nothing I’d chose above a Sun Solaris system. Would you? Honestly?
Put the power in the server where everyone can share the power, and give the $15kpa secretary a $100pa desktop, not a $2k desktop.
It’s about cost management, not just a new Linux distro.
MadHatter is a crap homr Linx distro, but an excellent IT strategy.
I respect SUN for what they are doing to Linux right now, they are leaving it alone, and letting open source development continue without conflict. I don’t think that Linux users are up to speed on open source, especially if they are criticizing SUN’s move.
Why is it that every time I read these posts there are a good number of individuals that are just amazed and shocked that someone like SUN would fund SCO… Well duh! Do you think that Sun wants Linux to take out its Solaris OS and jockey them out of the UNIX market. Why do you think Microsoft is funding them, simply because they want Linux to look bad when compared to Windows. This is business, not the super, happy, helping, hands competition of 2003. When other companies play hard ball, people expect it – when SUN does, everyone acts like it’s a personal affront. What SUN failed to realize is that a large portion of the Linux market is made up of Linuz zealots, choosing Linux for them is like a lifestyle choice, an aspect of who they are. As such, SUN spreading the usual competitive product FUD is taken out of context by such individuals, they translate attacks on Linux as being attacks on their beliefs or religion.
Also, regarding SUN’s ambivalent and painfully careful relationship with Linux. Solaris x86 is currently less than amazing, mainly due to limited hardware and software support. Yet, Linux is still behind Solaris in many respects. If SUN opened up Solaris x86 completely by open sourcing it, then they could potentially be free from the threat of Linux. But, they are tied up on Solaris x86, they can’t just open source it. This OS for certain has propreitary UNIX code and god knows how much copyrighted code. So they have to compete with Linux on their own turf, on x86 hardware. An unenviable position. But at least they are not under threat from SCO, they own the rights to their “System V” “derived” code.
However, then why would SUN sell their own Linux offerings? Well, SUN would be freaking stupid not to sell Linux to the countless individuals who want cheap x86 hardware with cheap “UNIX” a.k.a. Linux (which includes with it droves of software applications and the endorsement of every major IT publication out there.) This holds particularly true for the “neuvo” Linux admins that have never touched a proprietary UNIX and feel certain ethical and moral propriety to the cause of free Linux. But to these administrators, SUN is tainted, they don’t believe in the religion of Linux fully, they are not “true believers.”
What many Linux administrators fail to grasp is the Linux kernel is pretty good, but it is just another UNIX. From a technical standpoint, there is nothing revolutionary about Linux, if anything it is compelling because it is free with wide support. Solaris is more advanced than Linux on many points for the time being. This is already changing because of research and capital investments from several large companies which are looking towards Linux as the next de-facto UNIX standard. Not because it is better, but because a loyal following of zealots, IT trade magazines, and “CIOs” have decided that Linux is the uncontested replacement of high-priced, proprietary, UNIX.
These companies interested in Linux are smart, they realize the buzz that surrounds the OS and they want to be there to cash in on the Operating System when it does become the “standard” UNIX. However, when these large companies are done with it, Linux will be “Yet Another UNIX”. So the Linux zealots that are offering up their Linux to the alter of business, may well be shocked to find their Free Linux now costs them an arm and a leg and possibly their first-born (if they want a “certified” application running on systems that are “certified” by their hardware manufacturers and “certified” by RedHat/SuSE, with maintenance and updates, etc.)
So, should SUN then dump their worthy efforts on UNIX because of the irresistable force behind the Linux “movement”? Could you honestly say it would make sense for SUN to dump their Solaris x86 (a better Operating System in many respsects) in favor of Linux, just because fanatics assume that Linux is the best OS ever? Solaris represents years and years of research and development for SUN. If SUN promoted Linux over Solaris, then Solaris for SPARC would certainly lose validity due to SUNs Linux endorsement. An even more interesting question to ask may be, how many network admins would install Linux if they knew they could install Solaris x86 for the same price, with all the same software packages, fully supported on their hardware?
Linux has an edge over other Operating Systems not because of quality, but quantity. The quantity of binary compatible applications and the quantity of hardware systems that the OS can run on. If SUN could develop substantially more x86 drivers and encourage developers to port more applications for Solaris x86 (and maybe even create a great native version of Solaris for AMD64) then they may would have a chance of reviving a truly top notch operating system. If they aren’t up to this task, then no amount of posturing, finger pointing, or whining will make Solaris any less dead in a few years (Unless the SPARC ends up running circles around the PowerPC/Itanium and the price drops considerably, not a chance in Hades.)
So ultimately it comes down to this, if SUN execs would wake up and pull their heads out, they would realize they should be less concerned about making certain Solaris runs great on only SPARC, and rather make certain that Solaris runs on ANY hardware platform (in particular x86 and x86-64.) And they should make certain JAVA runs well on them, very well.
This is what I’ve been waiting to see. IBM making money on keeping Linux complex so they can push in loads of consultancy time. Now the serious player SUN enters the market and change all that.
I’ve had serious problem feeling confidence for the Linux players on the market, but SUN is a company I rely too when making decisions.
However, I still don’t understand why I’d want Linux over solaris as Solaris is indeed superior!
HAL (IP: —.adsl.hansenet.de) – Posted on 2003-09-27 20:04:58
raptor: *shrug* there isn’t enough trust there after the first interview to validate me putting any time and effort towards their linux products while other companies don’t change their public opinion on Linux once a week like Sun does.
Would it make any difference had Schwartz said:
“Also, let me really clear about our Linux server strategy. We don’t have one. We don’t at all. We do not believe that Linux plays a role on the server. Period. If you want to buy it, we will sell it to you. Our vision is Linux on the desktop, which is its primary strength and Solaris on the server, again, which is its primary strength. We see the two flavours of UNIX working in tadem to create an integrated end to end solution”
It’s because people are paranoid and ignorant, just like in a lot of KDE threads peopel say “Watch out, the Canopy Group, same group that owns SCO has a 5% stake in Trolltech, boycott KDE and Qt”. In reality, the Canopy group has stakes in dozens of companies and most of Trolltech is owned by Trolltech, this is just FUD.
Interesting that you bring up the Canopy Group because the CEO of it was the former Novell founder and CEO who Bill Gates dubbed, “the grumpy grandfather of IT”. Lets put it this way, the CEO of the Canopy has no love loss for Microsoft what so ever. He saw a company he built up get multilated by Microsoft overnight.
After he left Novell, he founded the Canopy Group. What we don’t know is how much pushing and shoving is done by the Canopy group towards SCO, or whether is is their own (SCO’s) plan. If you look at the amount of shares sold by the Canopy Group vs. how much the managers own, I would be more concerned about the activities of the management at SCO.
SCO ledgers show that two companies paid them money — Sun and Microsoft, both to the tune of several million dollars. This is the *only* money they made in the last quarter. I really do not believe all of this is as simple as wanting UNIXware drivers. Sun doesn’t need to get drivers from UNIXware. Solaris is different enough from UNIXware that it wouldn’t be a simple port. And the x86 *BSDs have enough of a driver base that, if they wanted to, Sun could just use those.
They need drivers written to the specifications of the hardware, not some ad-hoc development someone did in their basement based on some reverse engineered code. Unlike the BSD community, SUN has to back up their products through their limited liability. They can’t simply turn around and say, “oh, but we didn’t develop that driver, we simply ported it”.
SUN paid $8million for the required IP. That is chump-change. It would cost them a heck of alot more trying to port, test and enhancement BSD drivers.
Also, as a previous person said (in an interview), the money paid by Microsoft and SUN is lot a long term plan and if there was some alteria motive they would need a constant revenue stream rolling it, which isn’t happening.
I am wondering why SuSE allows Sun to build a distro based on its own product when Sun is targeting the same market as SuSE Linux Desktop.
SuSE is very bad in commercial strategy : since Sun is far bigger than SuSE, this company can wipe out SuSE Linux Desktop and take the monopoly on that market (Linux Desktop).
Strange, isn’t it ?
Also, as a previous person said (in an interview), the money paid by Microsoft and SUN is not a long term plan and if there was some alteria motive they would need a constant revenue stream rolling it, which isn’t happening.
So ultimately it comes down to this, if SUN execs would wake up and pull their heads out, they would realize they should be less concerned about making certain Solaris runs great on only SPARC, and rather make certain that Solaris runs on ANY hardware platform (in particular x86 and x86-64.) And they should make certain JAVA runs well on them, very well.
64-bit Solaris for Opteron coming, McNealy tells Reg
http://www.theregister.co.uk/content/3/30377.html
Linux on the desktop is the best match for Linux.
I were running a real business I would want tested certified hardware running tested certified software. That way I know my inventory system would run, my receivables wouldn’t error, and my company would ride through virus attacks. Too many people are of the mind that a real company can assemble a hardware software IT system out of parts and make it function 24/7 when they can’t. I know I have seen the ‘dead bodies’.
So a package of hardware and software makes perfect sense to the vast majority of companies who run businessses and don’t run techie playpens.
So Sun’s strategy of Linux(cheap) on the desktop and Solaris on the infrastructure makes more sense than anything any one else is offerring.
What is so difficult about BRANDING Sun’s Linux as JAVA that people don’t get? Do they know what branding is all about? IMHO it’s a brillant move, infers the desktop is for running JAVA which it is.
Bravo Sun for a brilliant strategy.
“What is so difficult about BRANDING Sun’s Linux as JAVA that people don’t get? Do they know what branding is all about? IMHO it’s a brilliant move, infers the desktop is for running JAVA which it is.”
What is confusing is that Java is a programming language. Now it seems to be a brand name for stuff written in C.
I think almost anyone would assume that a “Java Desktop” would be coded in Java, or at least the GUI would be.
It is going to be as confusing as calling a totally unrelated language “Javascript”.
What is so confusing about the branding? It would be like me questioning Windows XP and the “eXPerience” buzzword they tag to it. The question you bought up would be like me saying, “what am I meant to eXPerience? because I obviously haven’t eXPerienced anything yet”.
It is pretty basic. You have a server called the “Sun Java Enterprise System”, which users interact with via “Java Desktop System”, and for custom written applications (written by the ISP staff), then they recommend using, “Java Studio”.
To manage the whole lot easily and efficiently you then use what they dubbed, “N1”, which allows clustering of Solaris x86 and SPARC servers at the back end.
This whole package is called, “Sun Java System”.
(written by the ISP staff)
should be
(written by the IS staff)
By targeting Java (a business language) than you can integrate it into your IT infrastructure on Solaris (server side) or Linux (client side). This solves the question of ‘What langauge do I learn’? Just remember that Java is used by specialists for business solutions, it is not a generalist library for computer science research and development because in most cases you need a light weight (and broader) langauge for library development.
Here is another perspective. If you work for an organization who does not sell software, than use Java. On the other hand, if you work for an organization who only sells software, and that organization also performs software engineering (research and development) than learn Standard C or Standard C++.
I think that College undergraduates should focus on Java while graduate level studies should focus on Standard C and Standard C++. That would eliminate a lot of stress and it would simply life for many.
First of all, Sun’s offer is potentially a great cash saver for corporations: here you have a company that offers a full desktop operating environment, for the fraction of the cost of the alternative (WindowsXP + office). It’s an environment that has a well-integrated J2EE environment on top, so applications are easy to develop for it (a la Visual Basic, as someone notied).
It’s that simple.
Now, for Sun’s support of Linux: Sun is the only company not hindered by Microsoft in any way. I am surprised that Linux pundits always seem to forget that IBM and HP have deep interests with MS, and hence, you DON’T see them pushing it in areas where it would really compete against the mass-market king, Microsoft Windows. Sun can do it, and it is!
I think it’s just envy: opensource enthusiasts just can’t coincile with their fondest belief that a company can have an OS that is better than Linux and that’s still evolving. In fact, Sun might just replace Linux with Solaris x86 in the future, should they wish so. No problem whatsoever, and the winy Linux crowd would be left with one less reason to bark at the caravans that are moving along.
Solaris can not compete with Linux because one is controled by a vendor (Solaris) and the other is decentralized in terms of control by the open source community (Linux).
Linux is not going away, but Solaris almost fell out of the picture. Make no mistake, this deal by SUN was needed in order to resurrect Solaris from the dead.
I would like to see Linux develop non vendor infrastructure. I would rather see Linux focus on quality software and experimental infrastructure rather than standard vendor offerings.
BTW there most definately is ‘lock-in’regarding SUN’s desktop system, however a vendor might need to lock in customers in order to profit. If there was no lock in than maybe a vendor would go out of business. So let there be lock in, and the difference here is that lowering the prices to the degree that SUN is claiming (20 billion dollar industry to a 3 billion dollar industry) means that being locked in will not cost you an arm and a leg because there is enough volume (after cutting out middleware vendors).
>>Lets put it this way, the CEO of the Canopy has no love loss for Microsoft what so ever<<
Let me put it this way, the CEO of Canopy (which owns 45% of scox) is very much in bed with msft. Read scox’s last two 10Q’s if you don’t believe it. Msft has given scox $13 million over the last two quarters for a partial unix license – and msft doesn’t even sell unix. And msft might keep giving scox many more millions. Thanks to msft generousity, Yarrow is a *much* more wealthy man. Yarrow doesn’t hate msft so much he will turn away millions upon millions. Before msft became so generous, scox never had a profitable quarter, and was in danger of being delisted.
Linux is clearly a common enemy of msft and scox. And scox and canopy are essentially the same company. Do you know about the Vultus deal? Tell me how canopy isn’t pulling scox’s strings.
Let me put it this way, the CEO of Canopy (which owns 45% of scox) is very much in bed with msft. Read scox’s last two 10Q’s if you don’t believe it. Msft has given scox $13 million over the last two quarters for a partial unix license – and msft doesn’t even sell unix. And msft might keep giving scox many more millions. Thanks to msft generousity, Yarrow is a *much* more wealthy man. Yarrow doesn’t hate msft so much he will turn away millions upon millions. Before msft became so generous, scox never had a profitable quarter, and was in danger of being delisted.
That is SCO, not Canopy. Canopy is an investment company, not SCO itself. Seperate the two and get used to it. Just because Hathaway Investments has investments in cement production doesn’t mean that it is a cerment company.
Also, the IP bought by Microsoft is to enhance their UNIX offering which is pretty substandard when in comes to a UNIX API running on Windows NT. Also, the libraries are old, non-compatible with a number of UNIX specifications, the compiler is old and crappy and the speed is abismal.
Linux is clearly a common enemy of msft and scox. And scox and canopy are essentially the same company. Do you know about the Vultus deal? Tell me how canopy isn’t pulling scox’s strings.
“SCO WebFace Solution Suite, as the Vultus product will now be known, is a Web application development environment that allows customers to create and deploy applications in a browser without the need for installed plug-ins or Java. It is built on SCO’s Unix operating system, e-business services, and industry standards like XML, SOAP and UDDI.”
Oh, excitement, they bought an ASP vendor to bundle with their UNIXWare offering. Jeepers, what next, improving their product so that it compete? how dare they improve their product! they should listen to Walterbyrd and constantly whine and flog a dead horse on osnews.com
What should really be said here is that Solaris is superior to MS Windows. Solaris and MS Windows are products (not platforms…but middleware is a business platform, i.e. Java) and SUN and Microsoft are competing for business customers.
The Linux platform is a decentralized project. The fact that SUN offers Linux does not mean that SUN is implementing new features into Linux other than presentation, because it can’t do that…due to the GPL. It can’t turn the Linux platform into a product that forces users to upgrade. SUN poses no threat to SUSE or Redhat, the SUN distrubution only implements some presentation changes to GNOME 2.2. If SUN added features to the kernel, than they would have to be open source, and anyone could use the code. Solaris and Linux do not compete with each other, in fact, SUN has helped Linux avoid future entanglements with the law.
Remember that the state was develped to protect wealth. It allowed for the concentraton of wealth and power, so therefore entities with power can rely on the state to protect their interests.
Focus on Solaris versus MS Windows.
I would rather run the Java desktop system on a SUN Sparc platform workstation running Solaris rather than linux. Why not use real Unix if ya got it?
The Linux platform is a decentralized project. The fact that SUN offers Linux does not mean that SUN is implementing new features into Linux other than presentation, because it can’t do that…due to the GPL. It can’t turn the Linux platform into a product that forces users to upgrade. SUN poses no threat to SUSE or Redhat, the SUN distrubution only implements some presentation changes to GNOME 2.2. If SUN added features to the kernel, than they would have to be open source, and anyone could use the code. Solaris and Linux do not compete with each other, in fact, SUN has helped Linux avoid future entanglements with the law.
Correct, also, why is there a hatred of SUN simply due to the fact that SUN is letting the distributors make the decision about what direction Linux should head in. On one hand we had people complaining that IBM was trying to control Linux and being too heavy handed, yet on the other hand, when we have companies like SUN who take a more hands off a approach, they can sledged too.
What is it? do you want companies to be involved or do you want the community and distributors deciding the direction of Linux?
SUN has already embraced Linux as a desktop solution, isn’t that what most Linux advocates want, a large vendor willing to sell Linux as a desktop solution to the enterprise?
As for the cost of the software, on the server, $100 per-user, per year. Compare that to Redhat Linux Enterprise for $2499 + a per year maintance fee. That $2499 is PER SERVER, in the sun example, I could have 100 servers cluster together, and yet, I will only pay $100 per user accessing the server. That is alot better value for money than pay $2499 per server + maintainance contract.
I would rather run the Java desktop system on a SUN Sparc platform workstation running Solaris rather than linux. Why not use real Unix if ya got it?
So, are you going to say to Joe CEO cheapskate that they should spend AUS$3000 on a SUN Blade 150 instead of just AUS$1500 on a IBM ThinkCentre? How are you going to justify to this customer that it is worth their while spending AUS$3000 per-desktop? I would love to see your sale skills in action.
Sparc will come down in price if there is more volume, more people buying it…will drive down their production break even cost.
Linux for desktops.
Solaris for workstations/servers.
I don’t get why people claim Sun is sending “mixed messages” in regards to their Linux strategy. Clearly the Gnome 2.2 based MadHatter is much more polished than their Gnome 2.0 offering on the Solaris side, for any who have used either (which I’m guessing is a scant number of those commenting here, as the majority of people seem to be very much pro-Linux anti-Sun with nary a sign of real world Solaris experience)
This should be further emphasized by the fact that StarOffice 7 is seeing a Linux exclusive release in the short term, with a Solaris/SPARC release forthcoming.
“I would rather run the Java desktop system on a SUN Sparc platform workstation running Solaris rather than linux. Why not use real Unix if ya got it?”
Pricing aside, Solaris wasn’t designed for a workstation environment. It has some serious issues with interactive latencies at the console.
Solaris is a SERVER operating system. That’s where all of it’s strengths reside. Besides the problem of latencies, Solaris x86 and Solaris Sparc have next to no software or hardware support for desktop computing. It’s entirely geared to the server market. Outside of Apple and SGI, the only other hardware platform that has useable visualization systems (desktop or 3d modeling systems) are, you guessed it, x86 based PCs running MS Windows, Linux, or FreeBSD, NOT Solaris.
Unless Sun enacts a major rewrite at the kernel level Solaris x86 is not going to be a major contender in the desktop market. After the rewrite, they are going to have to give major incentives to desktop application people to create a new port for Solaris, an OS that has already died at least once. Ports aren’t cheap to create, especially if you are porting over Microsoft targeted applications. Sun has StarOffice, but StarOffice by itself is no match for the plethera of offerings in the Microsoft realm. They must also compete with the social inertia of “designed for Microsoft Windows” hardware. Purchasing a subset of SCO’s driver code isn’t going to help. SCO’s driver offerings are pathetically small and behind the times.
Sun either is going to have to go with the flow and allow Linux to do it’s thing, and at least APPEAR to enthusiatically support it, or try pushing Solaris x86 again to a mostly deaf audience. It’s one thing to offer what your customers want. It’s entirely another to appear unenthusiastic, or down right hostile to one of your own products. And yes, Sun Linux *is* a product, despite it’s open source-ness. No one in their right mind will buy a product that the company’s own executives disparage.
The first rule of salesmanship is: First impressions are extremely important.
The second rule of salesmanship is: Appearance is everything. Even if you have the greatest product in the world, if you or your representitives appear down on the product, or inept in your approach, it will never sell.
The third rule of salesmanship is: Listen to your customers and at least put in a convincing appearance of caring about what they care about.
Sun has dropped the ball on all three rules with both Solaris x86 and Linux. I see no reprieve in sight. In salesmanship, whathappens after the sale can be as important as the sale itself. Sun alienated a LOT of customers when they dropped Solaris x86. Saying “oops, we’re sorry!” doesn’t cut it when you have hundreds of thousands or millions invested in a certain platform, told your vendor is abandoning it, switch to a new platform, only to have the old vendor revive your original platform. That’s a good way to lose customers for good.
Sun alienated a LOT of customers when they dropped Solaris x86.
“customer” tends to imply someone who is actually giving you money. I have never heard any Solaris/x86 backlash from someone who has actually purchased anything from Sun, only home users/hobbyists/Linux zealots who have nothing better to do except complain about how a company who gave them an enterprise class operating system for free somehow owes them something more…
Sparc will come down in price if there is more volume, more people buying it…will drive down their production break even cost.
That is the chicken/egg senario. The fact remains that even if they were to get volume, SUN would NEVER be able to get their workstation prices down to something close to what Dell charges for their corporate desktop.
For their performance vs. price, even a Mac would be better value for money. The fact remains that no one in their right mind is going to deploy $3000 workstations for generic desktop use, it would be like a company deploying a whole heap of Apple Mac G5’s for generic desktop use.
Yes, I know that an average UNIX workstation from SUN lasts longer, in some cases up to 5 years, however, the simple fact remains that the average business “guru” never looks long term and aways demands instant gratification and savings even though in the long run their decision may cost the company more than a solution with a higher initial setup cost.
“customer” tends to imply someone who is actually giving you money. I have never heard any Solaris/x86 backlash from someone who has actually purchased anything from Sun, only home users/hobbyists/Linux zealots who have nothing better to do except complain about how a company who gave them an enterprise class operating system for free somehow owes them something more…
I think a good berometre for Solaris on x86 would be the Solaris x86 yahoo groups/mailing list. From what I have seen, there are a large number of people who have moved from Windows to Solaris for their server and are very happy with the server and software combo they received from SUN, also, a large number of these people are also using Solaris x86 as a desktop solution as well, and apart from some driver issues, again, most people are quite happy.
The only people who have problems with Solaris are people like Walterbyrd who jump onto osnews.com, flog that dead horse until it is a big pile of mince. Had his comments appeared anywhere else, they would be labelled as “flamebait”.
“Pricing aside, Solaris wasn’t designed for a workstation environment. It has some serious issues with interactive latencies at the console. ”
Excuse me I think Sun used to sell workstations before they did server. Trust me Solaris was designed with workstations also in mind.
What do you mean interactive latencies? Do you have any experimental proff to demonstrate said latency defficiency in the Solaris kernel?
It looks like you just looked at an article on the linux 2.6 kernel ranting about the O(1) scheduler and interactive latencies and concluded solaris probably is behind. Did you know solaris has had an O(1) scheduler for years now? Or that there is an interactive scheduling class (IA) to boost thread priorities for CDE/X/GUI threads. Also that Solaris has dynamically loadable schedulers with a Real Time scheduling class.
The Solaris kernel has been fully preemptable for years now.
Oh and the 1:1 thread model that the linux community is making a big fuss about has been in Solaris since 2000 default in Solaris 9 and introduced in 8.
” Unless Sun enacts a major rewrite at the kernel level Solaris x86 is not going to be a major contender in the desktop market. After the rewrite, they are going to have to give major incentives to desktop application people to create a new port for Solaris, an OS that has already died at least once. Ports aren’t cheap to create, especially if you are porting over Microsoft targeted applications.”
What are yu talking about? Rewrite what. The Soalris kernel has better loadable module support than linux has, you still can’t take any driver module for the 2.4.x series kernel and have it work drives compiled for 2.4.18 probably only work for a that release or maybe if lucky the previous one or the next. Good luck trying to compile a driver written for 2.4.18 on 2.4.20.
Go read Solaris internals before you go and talk about the Solaris kernel. The linux kernel is premitive compared to Solaris in more ways than one. Linux has also borrowed quite a bit from Solaris. The slab allocator comes to mind.
It looks like you just looked at an article on the linux 2.6 kernel ranting about the O(1) scheduler and interactive latencies and concluded solaris probably is behind. Did you know solaris has had an O(1) scheduler for years now?
Thank you for mentioning this. Most Linux zealots that troll OSnews have no exposure to Solaris’s scheduling, specifically that Solaris allows multiple schedulers to be running simultaneously, that you can move processes between schedulers, dynamically load and unload schedulers, etc. whereas Linux has only one scheduler, and as you mentioned Solaris’s timeshare scheduler is O(1) and has been for quite some time.
That said, what was being argued in the original post is true: Solaris is a high latency, high throughput operating system, whereas Linux is a low latency, low throughput operating system. Several times in the past I have pasted various Linux vs. FreeBSD vs. Solaris benchmarks using a number of both disk and network benchmarking tools including dbench and rawio. In all cases, Linux was the throughput loser, with Linux 2.6.0-test4 scoring significantly than FreeBSD 5.1 and Solaris 9 x86 on the same hardware. The problem with optimizing system call algorithms for quick execution (and thus low latency) is that they tend to be much less efficient than more complicated (and thus high latency) algorithms designed to optimize throughput.
Recent Linux kernel design has far too much of a desktop bias to perform well in the server arena compared to operating systems which have been consistently optimizing for server instead of desktop use. I’ve read several messages on Linux kernel mailing lists warning that perhaps Linux is trying to fix problems which should be addressed in userspace in the kernel too often which are going unheeded. Windows and MacOS X have both addressed this problem by prioritizing the process which owns the currently focused window, as an example of a userspace fix.
For those curious, here are my old dbench [ http://samba.org/ftp/tridge/dbench/ ] numbers. I can go into this in more detail if anyone is curious but it grows quite tedious to do this in every post where Linux zealots are arguing that their low kernel latency makes for a better kernel in general, rather than realizing that optimizing for low latency has horrible effects upon throughput:
Linux 2.4.20 (XFS):
Throughput 17.5915 MB/sec (NB=21.9893 MB/sec 175.915 MBit/sec) 64 procs
Linux 2.6.0-test4 (ext3):
Throughput 23.8191 MB/sec (NB=31.0049 MB/sec 250.041 MBit/sec) 64 procs
Solaris 9 x86:
Throughput 29.2046 MB/sec (NB=36.0637 MB/sec 288.510 MBit/sec) 64 procs
FreeBSD:
Throughput 31.9033 MB/sec (NB=39.8792 MB/sec 319.033 MBit/sec) 64 procs
Thanks for pointing that benchmark out. I have to agree with you that solaris has favoured high throughput massively timeshared systems but it also has the IA scheduling class. This class was added to benifit the OpenWindows, CDE and Xwindows systems, where by the current window under focus gets a priority boost over other threads. All process spawned off under The GUI inherit the IA schduling class.
Solaris by default is designed for SMP systems. It looks like there was an article about the linux 2.6-2.4 kernel bench marks, in which linux 2.6 with SMP scalability improvments performs worse than the 2.4 series for single cpu systems.
It looks like SMP scalability has a slight penalty for in terms of latency of acquiring locks. Solaris has always favoured scalabily at the cost of a little performance. It looks like the linux developers are having to make the same compromises. While I applaud the linux developers for thier hardwork and effort it perturbs me that some linux users read articles, pick up on jargon and make baseless claims.
Solaris by default is designed for SMP systems. It looks like there was an article about the linux 2.6-2.4 kernel bench marks, in which linux 2.6 with SMP scalability improvments performs worse than the 2.4 series for single cpu systems.
True, I have seen Solaris run on a SMP configuration (Dual Xeon) and you really do notice the difference as you add more processors to the machine.
It looks like SMP scalability has a slight penalty for in terms of latency of acquiring locks. Solaris has always favoured scalabily at the cost of a little performance. It looks like the linux developers are having to make the same compromises. While I applaud the linux developers for thier hardwork and effort it perturbs me that some linux users read articles, pick up on jargon and make baseless claims.
If you look at the price of SMP configured machines, I wouldn’t be surprised if in the future SMP configurations will be a pretty normal site on peoples desktop. Single CPU server shipments will soon be outsold by SMP machines and ultimately this is where highly tune operating systems like Solaris will kick in.
Opteron, which Solaris 64bit is going to be ported to, will really benefit when 8way configurations are sold.
My point mainly come from working as a HP-UX SA in a large environment 50+ servers. The team of seven that handled the HP environment had an HP Workstation $5000 for remote X11 and different terminals along with having a machince that represented the environment for testing etc.. and a PC $1000-$5000 mainly for just for Exchange email services and web surfing. CDE works but is bare bones plain. Sure you can down load the HP-UX version of gnome. But wouldn’t be cool if you could install a whole layer of extended OS features to include apps like ximian (to plug into exchange) mozilla , star office etc.. on top of Sparc solaris or HP-UX the primary benifit that your workstation OS would not deviate from the environment you work and would have all the functionality of the windows box. Truth is the old UNIX dogs i worked with very very talented people both HP and SUN sa’s have no interest in Linux it is still to them a toy OS(sorry i tend to agree also even though i really like linux) and a liability becasue the source is freely available. These are the people that will be IT managers in the coming years and its these guys that you have to convinve open source is a good idea. Do that by Open Source software first Operating system later.
>>So Sun’s strategy of Linux(cheap) on the desktop and Solaris on the infrastructure makes more sense than anything any one else is offerring. <<
So where does x86-Solaris fit into that strategy? Certainly sun is not advocating x86 for servers. But, if linux is the desktop strategy, why spend all those millions on developing x86?
In the last two qurarters sun has given scox $5 million for x86 drivers. This is especially interesting since, from everything I have ever read, scox has misserable hardware support for x86.
>>Linux woke Sun up about its semi-dormant x86 Solaris, which was mainly lacking in drivers for current x86 hardware <<
Why now? Linux has been around ten years. But it wasn’t until scox started their anti-linux campaign that sunw “woke up.” Interesting timing to say the least.
>>Also, as a previous person said (in an interview), the money paid by Microsoft and SUN is not a long term plan and if there was some alteria motive they would need a constant revenue stream rolling it, which isn’t happening.<<
How do you know? Neither msft, or sunw, has claimed that they will stop funding scox. For all we know, msft and sunw will be giving scox even bigger contributions through 2004.
I do hate posting on this cesspool, but someone needs to correct the missinformation. =/
Linux is a low latency, low throughput operating system.
Not true. There are options for reducing latency, but these are selectable at compile time. Everyone concerned is very aware of the potential throughput/latency tradeoff – which is why lots of things (eg. preempt, elevator algorithms) are selectable.
Several times in the past I have pasted various Linux vs. FreeBSD vs. Solaris benchmarks using a number of both disk and network benchmarking tools including dbench and rawio. In all cases, Linux was the throughput loser, with Linux 2.6.0-test4 scoring significantly than FreeBSD 5.1 and Solaris 9 x86 on the same hardware.
Yeah, well, you’ve also posted that the same box takes 3 seconds to launch javac (hot cache). Pardon me if I disregard your results.
Here are my dbench results. I’ve chosen hardware which is, I think, comparable to yours. Box is an Athlon MP1800+, running 2.6.0-test5-mm4 with a 40GiB 5400RPM IDE drive and 512MiB of RAM (actually has 4GiB, but I booted with mem=512M). Tests done on a ReiserFS3 partition. GDM and a few services are running in the background, wiring ~100MiB in total.
Anticipatory IO scheduler:
Throughput 48.5834 MB/sec (NB=60.7293 MB/sec 485.834 MBit/sec) 64 procs
Deadline IO scheduler:
Throughput 45.5236 MB/sec (NB=56.9046 MB/sec 455.236 MBit/sec) 64 procs
The problem with optimizing system call algorithms for quick execution (and thus low latency) is that they tend to be much less efficient than more complicated (and thus high latency) algorithms designed to optimize throughput.
[/i]
That is a very risky generalisation to make. There are a number of instances where reducing algorithmic complexity increases throughput and decreases latency. The scheduler in 2.6 has a much lower latency than in 2.4, but it’s also got a much, much higher throughput.
That isn’t the only, or indeed main, kind of latency optimisation being undertaken. The biggest would be the addition of kernel preemption, which is a compile time option. There is also an IO scheduler designed to reduce latency, but that’s an option as well.
Recent Linux kernel design has far too much of a desktop bias to perform well in the server arena compared to operating systems which have been consistently optimizing for server instead of desktop use. I’ve read several messages on Linux kernel mailing lists warning that perhaps Linux is trying to fix problems which should be addressed in userspace in the kernel too often which are going unheeded.
OMG bascule has read some messages on LKML. STOP THE PRESSES!
Give us a break. You seem to know a bit about FreeBSD and Solaris, but your posts of Linux border on outright lies. The improvements in scheduling, threading, locking, large memory support and block IO don’t benefit my laptop much. They do benefit my 8 processor server and it’s brethren.
There are too many involved parties to allow Linux to become “over optimised” for desktop use.
walterbyrd (IP: —.59.16.246.Dial1.Denver1.Level3.net) – Posted on 2003-09-29 14:05:35
>>Linux woke Sun up about its semi-dormant x86 Solaris, which was mainly lacking in drivers for current x86 hardware <<
Why now? Linux has been around ten years. But it wasn’t until scox started their anti-linux campaign that sunw “woke up.” Interesting timing to say the least.
Linux has only started to become reasonable on the server in the last 4 years. How long have you been using it for? 1-2 years because the media has jumped on the hype? listen to be sonny, I’ve been using it for 9 years, so get down off your high horse and start listening to those who have a clue.
walterbyrd (IP: —.59.16.246.Dial1.Denver1.Level3.net) – Posted on 2003-09-29 14:01:46
>>So Sun’s strategy of Linux(cheap) on the desktop and Solaris on the infrastructure makes more sense than anything any one else is offerring. <<
So where does x86-Solaris fit into that strategy? Certainly sun is not advocating x86 for servers. But, if linux is the desktop strategy, why spend all those millions on developing x86?
In the last two qurarters sun has given scox $5 million for x86 drivers. This is especially interesting since, from everything I have ever read, scox has misserable hardware support for x86.
SUN paid $8 to extend their current UNIX license and also to license drivers and other x86 related parts of UNIXWare, the second part of the agreement was an offer to purchase shares below the current market price. SUN hasn’t taken them up on that offer.
If SUN did have a vested interest in SCO, they would have bought those shares when they had the chance. They have $5.7 in the bank, it isn’t as though they are running out of money very soon.
Also, that $8million was a one off payment. There have been no transactions between SCO and SUN since then.
Also, I find it funny that you do a lot of “reading” and no experiencing. OpenServer and UnixWare has way better hardware support than Solaris x86. Solaris x86 is close to what FreeBSD was back in the 3.x series.
As for Solaris x86 on the server, why not? they sell X86 servers, why not load it with their own operating system? How is that any different to HP selling HP-UX equiped Itanium and PA-RISC servers and workstations?
They have $5.7billion in the bank, it isn’t as though they are running out of money very soon.
First of all, it appears that you are equating workstations or desktops with having more than one CPU. That’s a rare situation. Workstations generally only have a multiple CPU base when there is a need to prove concepts on small SMP systems, or the graphics designer/CAD technician that wants a little more computational power from his or her machine. You call me a Linux “zealot” which I’m far from being. I don’t care what OS people use so long as it fits their need. What you are ignoring is that the vast majority of desktop computers and workstations are single user systems with hardware designed for x86 based motherboards, and only have a single CPU even in an industrial environment. Sure Solaris is an awsome OS for servers, I will not debate the point. But an awsome server OS makes for a poor desktop OS compared to an OS that was designed to handle the different needs of the desktop user. I’ve used and administrated Sun, IBM, SGI, x86, and older DEC systems, However, Sun still performs poorly as a standard single user desktop. Sun is not among my choices and never will be for the desktop arena because of lack of hardware support, and the SMP bias of it’s Solaris OS.
People here quote study after study on performance, initial price, TCO, etc etc. (“64 proc system”?! That’s a study that has zero bearing on desktop performance!) But all of the conclusions boil down to personal bias for or against. For my own purposes, and those of the companies I’ve worked for, Solaris just does not have adequate desktop/workstation support in either software applications nor hardware devices to be considered a viable alternative. The vast majority of the companies I’ve worked for are MS desktop, MS groupware, and IBM AS/ series for database servers. A few were Sun Sparc server, and none were Solaris desktop either x86 nor Sparc. One used an SGI Crimson and an Onyx for number crunching and the workstations were UP PPC systems.
Yes Linux does have issues on server throughput, it has issues with desktop throughput as well. The linux kernel designers are trying to have their cake and eat it too. You can’t have a server OS and a desktop OS in the same code tree that performs as well as a OS that was primarily designed for one or the other. Either you will have a good server OS that performs mediocre on the desktop, or you will have a desktop OS that performs poorly as a server, or worse, a mediocre desktop OS that’s also mediocre as a server. This supposedly “linux zealot” is one that thinks Linux is mediocre at both desktops and servers, in both upcoming 2.6 and current 2.4. There is a vast difference in the requirements between SMP and UP systems as far as optimized performance goes, and they aren’t fully compatible in the method of addressing their requirments.
The reason I think Linux v 2.6 is mediocre has nothing to do with it’s performance, it’s entirely a driver support issue compared to the only other real alternative, Microsoft Windows. 2.6 scheduling for the desktop has made a near quantum leap in responsivness of the desktop when coupled with the multimedia tree and low latency options. However, when you compare Linux hardware support that’s directly supported by the hardware vendors versus vendors that directly support Microsoft (and even Apple) Linux hardware support becomes mediocre at best. But like I said, Linux is still not “there” when it comes to very high end servers, due in part to the new threading system someone else mentioned. But that new threading system does NOT benefit desktop users enough to be a championed feature! This is not about Sun’s Solaris viability as a server platform, it’s about it’s viability as a desktop, and as such it has zilch in most situations.
“But like I said, Linux is still not “there” when it comes to very high end servers, due in part to the new threading system someone else mentioned. ”
This should be “But like I said, Linux is still quite not “there” when it comes to very high end servers, despite the new threading system someone else mentioned.”
>>That is SCO, not Canopy. Canopy is an investment company, not SCO itself.<<
Again, canopy owns 45% of scox – repeat 45%. And that is only the direct ownership. Canopy companies also own shares in scox: 25 canopy companies own about 1.5% each, that’s about another 37% of scox that canopy controls indirectly. Why do you think ibm has issued a sopena to Yarro? Canopy controls scox.
Apparently, you don’t know anything about the vultus deal. For example, did you know that vultus and scox are both canopy companies, and that both were located in the same building, and that building was owned by canopy? Did you know that vultus was “bought” entirely with scox shares, and that all the money went to canopy? Did you know that vultus products only work with MSIE – not with scox software? Did you know that about 10 of the 25 canopy companies have the same address – I mean even the same suite number? Maybe you don’t understand the way canopy opperates.
>>Also, the IP bought by Microsoft is to enhance their UNIX offering<<
Msft doesn’t have a unix offering and never will. It’s pure fud money. Question: why now? Why all of the sudden? Isn’t it interesting that sunw and msft all-of-the-sudden *need* scox at the exact same time that scox launches it’s campaign to pirate linux? What an amazing coincidence.
BTW: you have been a great scox supporter, correct? Are you still?
>>why is there a hatred of SUN simply due to the fact that SUN is letting the distributors make the decision about what direction Linux should head in. <<
That is *not* the reason for the hatred of sun, not even close. The reasons for the hatred of sune are:
1) Sunw is clearly support scox. Scox is desperately trying to invalidate the GPL, and take full ownership of Linux. According to scox, scox and only scox controls linux. According to scox, unless you buy your linux from sunw, you own scox $699 per CPU.
2) In return for sun’s support, sun is allowed to claim that only sun’s version of linux is legal, and if you buy from anybody else: scox will sue you.
Sun is pulling a scam, that is why people who used to like sun, have now turned against sun.
>>Linux has only started to become reasonable on the server in the last 4 years. How long have you been using it for? 1-2 years because the media has jumped on the hype? listen to be sonny, I’ve been using it for 9 years, so get down off your high horse and start listening to those who have a clue. <<
On what are basing these assumptions? What makes you think I use linux at all? And, if you are going to jump to that assumptions, why assume that I have only used linux for 1-2 years? You are jumping to conclusions all over the place. Why do you claim I am on a “high horse?”
>>If SUN did have a vested interest in SCO, they would have bought those shares when they had the chance.<<
I don’t think sun can excercise warrants until May 2004.
>>They have $5.7 in the bank, it isn’t as though they are running out of money very soon.<<
No, but it’s not a normal “arm’s length” transaction.
>>Also, that $8million was a one off payment. There have been no transactions between SCO and SUN since then. <<
Two payments so far. Could be more to come.
>>Also, I find it funny that you do a lot of “reading” and no experiencing. OpenServer and UnixWare has way better hardware support than Solaris x86. Solaris x86 is close to what FreeBSD was back in the 3.x series. <<
Lots of assumptions on your part, again. UnixWare may have better hardware support than x86-solaris, but still far inferior to linux.
>>As for Solaris x86 on the server, why not?<<
I have never know sunw to advocate 32-bit x86 for anything. Seems to me Sun has been bashing that platform for years.
Not true. There are options for reducing latency, but these are selectable at compile time. Everyone concerned is very aware of the potential throughput/latency tradeoff – which is why lots of things (eg. preempt, elevator algorithms) are selectable.
You’re entirely correct, and I did not explicitly disable the low latency settings in the kernel for that benchmark. I was trying to do some “out of box” benchmarks with as vanilla of options as possible.
Here are my dbench results. I’ve chosen hardware which is, I think, comparable to yours.
I really need to rerun all my benchmarks anyway, but unfortunately I don’t have any dual boot systems anymore. I’d also like to find something better than dbench, which doesn’t seem designed to build on anything but Linux and requires considerable hand tweaking to build on FreeBSD/Solaris. There’s also been complaints about dbench’s usefulness as a benchmark, such as http://mail.nl.linux.org/linux-mm/2001-07/msg00126.html
As for your numbers, it makes me wonder if something is wrong with my setup under Linux, possibly the drivers for my SCSI host adapter (Adaptec 29160). I’ve been using the Adaptec supplied ones rather than the one crossported from FreeBSD years ago. I should also give ReiserFS a try in the benchmarks as its performance characteristics might also lend some help to Linux’s scores.
That is a very risky generalisation to make. There are a number of instances where reducing algorithmic complexity increases throughput and decreases latency. The scheduler in 2.6 has a much lower latency than in 2.4, but it’s also got a much, much higher throughput.
This is an odd example to give, considering the stalling bugs the O(1) scheduler was experiencing for quite some time. However, that aside it is generally true that optimizing for low latency requires throughput to go by the wayside. As it is a generalization I tried to back it up with benchmarks.
OMG bascule has read some messages on LKML. STOP THE PRESSES! Give us a break. You seem to know a bit about FreeBSD and Solaris, but your posts of Linux border on outright lies.
Can we please stop with the personal insults? Thanks…
I just see far too much ignorant mudslinging on these forums, and the only means I see of quantifying exactly how operating systems compare is through benchmarks. I’m quite glad you posted some dbench numbers, and it would be great if you could get some more numbers for operating systems other than Linux on the same hardware.
I was trying to do some “out of box” benchmarks with as vanilla of options as possible.
There isn’t much value in that. The kernel being shipped in RHAS isn’t going to have the same options enabled as the kernel shipped with Lindows. The same applies to other OS’s – choose the options likely to be used in production.
I should also give ReiserFS a try in the benchmarks as its performance characteristics might also lend some help to Linux’s scores.
The difference between ext3 and reiserfs was marginal when I tried. IME on larger, multi spindle systems ext3 in 2.6 can be quite a bit more scalable than reiserfs3.
This is an odd example to give, considering the stalling bugs the O(1) scheduler was experiencing for quite some time.
Not really.
a) bugs being the operative word
b) they mainly manifested themselves with interactive processes
and if you dimwits would have read the interview with Yaro, who owns the Canopy Group, you’d know that this guy, and by that Canopy, pushes the SCO cause and backs them fully. That again means that any of the companies Canopy has interest in are compromised.