Sun has recently released Solaris 10. It is currently free, as in beer, and most of it is promised to be released under an OSI approved license in the second quarter of 2005. Most everyone reading this probably knows all of that. The release and subsequent open sourcing of Solaris 10 has caused quite an uproar in the Open Source community and the IT industry as a whole. Read the review here.
You might want to read this from an article on EnterpriseLinux:
http://www.enterpriselinux.org/stories.php?story=05/02/17/9649385
“Building on eight-way Linux support it introduced in August 2004, HP is the first major vendor to scale Linux to 64-way on a standard Linux kernel through a project code-named “Big Tux.” This will help eliminate the risks associated with custom kernel implementations, such as those offered by competing firms.”
So Linux has been able to scale beyond 8 CPU’s since when? It looks to me like a relatively recent development and it certainly doesn’t look like a mature technology by any sense.
You just don’t get it do you? Can I buy a version of Linux that supports more than 8 CPU’s out of the box? I do not mean “sign the NDA and give us several million dollars”, and not “well we support large CPU configurations in clusters”, can we run it on one box with a large number of processors. The answer is NO, and unixconsole is right. And just because it works in the lab, does not mean it works on a production workload.
The reason why you consistently get slammed here is you attittude, which for lack of a better word, sucks! Who died and appointed you the “keeper of all things Linux”. Maybe if you toned down your rhetoric from “All I’ve been trying to do is dispute the misconceptions/lies/whatever which Solaris fanboys have been dishing out.” to something more professional, maybe people would be more willing to listen to your point. Instead it comes out as “yet another whining Linux zealot” pitching a fit because someone had the nerve to say something bad about “my favorite OS”. You complain about anyone saying you are a Linux fanboy, yet use the same terminology on Solaris users without skipping a beat “Are Solaris fanboys really that afraid of Linux getting into the “last bastion of Sun”, aka, big iron?”. Talk about two faced! If you want respect, you are going to have to earn it, and right now you don’t have mine.
Now here is my rant, I have used Linux off and on since 1993 (before all the pretty, easy to use GUI configuration nonsense). I was using Linux when it was more user unfriendly than regular Unix! I see Linux as a solution looking for a problem. What can Linux do that Solaris, AIX, and HP-UX can’t, from what I see not much. An operating system is a tool, just like the computer is. And as there are multiple operating systems, there are multiple choices. Not everybody thinks Linux is the solution for every problem, although that doesn’t stop people from trying to bend it into something that will. Those of us who use and recommend Solaris, do so because we are familiar with the product and what it will and will not do in a given situation. For some situations I recommend Windows, some Solaris, some AIX, and some Linux. It all depends on a variety of factors, and “cool points” is not one of them.
If you have something relevant to add other than your stance that “Linux is better because you say so” fine, if not don’t comment.
Are you going to say that Linux doesn’t run my computer because the nVidia GPU, the Northbridge, the Ethernet adapter, etc all have embedded OSs in ROM or ASICs? Having an ASIC or ROM with an embedded OS to control dedicated hardware is very common when a full blown OS is overkill.
I really think you should stop digging a deeper hole here. You have a very limited understanding of compters and your statements above proves it. The only embedded code any of those ASICs have is fcode for open firmware based system or boot code. Sorry they are not Oses.
But you have done nothing to dispute Linux’s ability to run on large system. If you’d like another example, here’s an IBM system, http://www-1.ibm.com/servers/eserver/pseries/hardware/highend/595.h…… That’s an up to 64 way Power5 system which runs AIX or Linux. If you look at their Performance data page, you can see them specifically talking about a 32 way Linux system. They don’t appear to specifically mention any other Linux configuration.
Read the fine print in the above pasted URL.
Support for Linux running in single system image mode (non-LPAR) requires a network attached pSeries system with AIX 5LTM. Linux running in an LPAR on the p655 requires either AIX 5L in an LPAR on the p655 or a network attached pSeries system with AIX 5L.
You can’t run linux on a 32 way box without also buying and running AIX.
I am not going to address the rest of your post becuase you have absolutely no clue.
Admit you were wrong and move on,
Linux isn’t run on the compute nodes because it’d be a waste, the compute nodes don’t need a full blown OS. And the compute nodes “do the real work” in the same way a GPU does the real work of rendering graphics. Without the “management node”, aka the CPU and OS, the GPU is useless. The same is true of the Bluegene architecture. And Linux is, according to that article, still running on 1024 of the processors. Even without the compute nodes, the Linux part of Bluegene is bigger than the Sun system you pointed out.
One the blueGene linux on the I/O node is also a stripped down kernel where no userland code runs only I/O from the computed nodes. I wonder why they didn’t just use embedded linux for the compute nodes.
Linux is running on a 1024 way cluster of I/O nodes. Repeast after me “A Cluster is not a SSI smp machine”. Repeat it till you get it and then some.
The bottom line is you can’t use BlueGene as an example for linux scalability. Almost any linux fanboy but you will get it.
Also I have proven that there is always a fine print when it comes to linux running on big iron. And all the examples you have given have had that fine print. There is no fine print while running any of the other OSes.
IBM and HP are using linux as a bargaining chip to push their machines. A linux license on a pSeries also sells an AIX license or worse an entire pSeries running AIX.
If that doesn’t enlighten you to linux’s state in the enterprise today, you linux fanboys need to buy a clue stick and whack yourselves on the head with it.
Your technical arguments are very weak. Your attitude is very poor. Please stop.
Very well said!
So does Linux scale well or not? This ain’t stopping until the Solaris fanboys admit their blatant mistakes?
I guess the true Solaris fanboys have really shown their colors.
You can’t run linux on a 32 way box without also buying and running AIX.
Per your own admission, Linux runs on a 32 way systems. The fact that it’s using AIX to bootstrap the system is irrelvant, Linux runs on a 32 way system. Per your own admission in a prior post, Linux runs on a 16 way partition on an HP box. But you’re still arguing.
The only embedded code any of those ASICs have is fcode for open firmware based system or boot code.
You might want to do a bit more research. Your video card and ethernet adapter have a ton of firmware on them. No, they’re not full blown OSes, yes, that’s the entire point I was trying to make. And, honestly, do you really think you are convincing anyone other than other Solaris fanboys unable to swallow the truth at this point?
One the blueGene linux on the I/O node is also a stripped down kernel where no userland code runs only I/O from the computed nodes. I wonder why they didn’t just use embedded linux for the compute nodes.
Google does the same thing on their cluster, are you going to claim they don’t run Linux? And the reason they didn’t use an embedded Linux is because what’s running on the compute nodes hardly qualifies as an OS. It has just enough code to bootstrap the processor, and to start a single process running.
Also I have proven that there is always a fine print when it comes to linux running on big iron. And all the examples you have given have had that fine print. There is no fine print while running any of the other OSes.
So now your argument is that Linux runs on that hardware but there’s fine print? Doesn’t that mean that Linux runs on systems with more than 8 CPUs? Doesn’t that mean that unixconsole was wrong? Are you going to admit it yet, though? Probably not.
Your technical arguments are very weak. Your attitude is very poor. Please stop.
You’ve yet to offer any technical arguments. All you’ve done is sit there and flail away after you repeatedly admitted yourself that Linux scales past 8 CPUs. Wouldn’t your attitude start to get pretty poor if after showing someone something for the 10th time the person is still unwilling to accept it? http://www.tpc.org/tpcc/results/tpcc_result_detail.asp?id=104062801 There’s an NEC machine running SuSE linux on 32 Itanium processors. I can keep finding systems if you’d like. I’m not exactly sure what’re you’re ultimately look for since you seem completely unable to just come out and say that Linux scales past 8 CPUs.
You just don’t get it do you? Can I buy a version of Linux that supports more than 8 CPU’s out of the box? I do not mean “sign the NDA and give us several million dollars”, and not “well we support large CPU configurations in clusters”, can we run it on one box with a large number of processors. The answer is NO, and unixconsole is right. And just because it works in the lab, does not mean it works on a production workload.
You just don’t get it. HP sells a system which supports an nPar of 16 processors on Linux. IBM sells a system with data of up to a 32 way Power 5. NEC has submitted TPC data for a 32 way Itanium for whic there is lots of marketting material on their site, but no “buy now” button. So I don’t know if that one is shipping right now or not. I’m showing you systems you can purchase that support over 8 CPUs. That means unixconsole is wrong. And if you look at the 3 systems I’ve pointed out (so far, I’m sure I’ll need to dig up more before this is over) they all run SuSE Enterprise Linux 9 or Red Hat Enterprise Linux 3. It looks like anyone can buy them from Novel or Red Hat without an NDA. Go ahead and look at the full disclosure for the NEC system on the TPC site, it even has the cost of the SuSE licenses … $300 and something .. same as retail. Red Hat will even give it to you for free for 30 days. Buying the hardware will cost you millions of dollars though, but you also said no one would buy Linux on million dollar hardware. I’m pretty much 100% sure you won’t admit you’re wrong about that either though.
The reason why you consistently get slammed here is you attittude, which for lack of a better word, sucks! Who died and appointed you the “keeper of all things Linux”. Maybe if you toned down your rhetoric from “All I’ve been trying to do is dispute the misconceptions/lies/whatever which Solaris fanboys have been dishing out.” to something more professional, maybe people would be more willing to listen to your point.
I’m sorry, how many posts in did I start that rhetoric? I mean, you went out of your way to insult me when I pointed out (correctly) that you can get Linux EAL 4 certified. Why should I be polite to people who after a series of posts have shown themselves to be nothing but fanboys that can’t do their own research AND STILL WON’T ADMIT THAT LINUX SCALES PAST 8 CPUS!
Instead it comes out as “yet another whining Linux zealot” pitching a fit because someone had the nerve to say something bad about “my favorite OS”. You complain about anyone saying you are a Linux fanboy, yet use the same terminology on Solaris users without skipping a beat “Are Solaris fanboys really that afraid of Linux getting into the “last bastion of Sun”, aka, big iron?”.
ROFL. You can’t be serious.
a) Someone didn’t say something “bad” about Linux, they said something which isn’t true. If it was 2002 and you said Linux can’t scale past 8 CPUs I wouldn’t have said a word because it couldn’t. Today it can. But you still refuse to admit that. Which leads to.
b) I think my Solaris fanboy comment is right on target at this point. All any of you have needed to do to get me to stop is to admit that there is ample evidence of lab tests and shipping systems that Linux can scale past 8 CPUs. So far, not one of you fanboys has been willing to meet the challenge of being honest. I’ve been honest about my mistakes so far. And, as far as I can tell, I’ve been beating the same drum this entire time. If telling someone that they’re wrong and here’s the evidence to prove it is all it takes to be a fanboy in your eyes, though, then where do I sign up?
If you have something relevant to add other than your stance that “Linux is better because you say so” fine, if not don’t comment.
See, this is where I don’t understand why all the Solaris fanboys are coming out of the woodword with their panties in a twist. Point out to me where I said Linux is better than anything. The closest I got to that was when raptor started a pissing contest about supercomputers. And even then I didn’t say Solaris couldn’t run on super computers, I think you’re the one that said it wasn’t designed for it. I just said, don’t start a pissing contest because Linux runs the biggest super computers in the world. (Note to everyone, you’ll see how as soon as I point out Bluegene, he attempts to show how Linux only runs some of it. A very telling personality trait I would say). Point out to me where I have said anything derogatory about Solaris or any other OS.
See, the difference between you and me is that you have been lying about Linux’s capabilities this entire thread. I have tried as best I can to not lie about any system. I’ve admitted I was wrong when I was. I’ve bent over backwards trying to provide evidence. So is the only excuse you have left for not admitting that Linux can scale past 8 CPUs at this point is that I “lost your respect”?
Per your own admission, Linux runs on a 32 way systems. The fact that it’s using AIX to bootstrap the system is irrelvant, Linux runs on a 32 way system. Per your own admission in a prior post, Linux runs on a 16 way partition on an HP box. But you’re still arguing.
No I admitted to nothing. I just merely pointed you to the fine print, that accompanied your URL.
If Hp lied about the standard distribution and IBM has fine print on linux, I wouldn’t be too quick to accpet what they say in thier PR and marketing literature. You have yet to give me irefutable facts being as I have refuted all of your posted links.
You might want to do a bit more research. Your video card and ethernet adapter have a ton of firmware on them. No, they’re not full blown OSes, yes, that’s the entire point I was trying to make. And, honestly, do you really think you are convincing anyone other than other Solaris fanboys unable to swallow the truth at this point?
I don’t have to do any research, I develop kernel and device driver software and know how this stuff works. If you believe that there are embeded OSes (frimware) that actually makes you GPU or ethernet interface work, I have bad news for you about Santa Claus and the Tooth Fairy.
The only device I know of that has embedded firmware is LSIs new line of SCSI controller and even those need DMA services and a device driver to work.
Obviously you don’t have he technical skill to understand this stuff, so I am not going to bother with further discussions with you.
Google does the same thing on their cluster, are you going to claim they don’t run Linux? And the reason they didn’t use an embedded Linux is because what’s running on the compute nodes hardly qualifies as an OS. It has just enough code to bootstrap the processor, and to start a single process running.
Google does what BlueGene’s architecture does???? Repeat after me “A cluster is not a SSI smp box”. Go read up on Googles architecture and then we can talk more.
So now your argument is that Linux runs on that hardware but there’s fine print? Doesn’t that mean that Linux runs on systems with more than 8 CPUs? Doesn’t that mean that unixconsole was wrong? Are you going to admit it yet, though? Probably not.
My point is everyone of your Linux scales to 64 way cpus has been refuted by some nice fine print. The thread between you and me was on HP’s article on linux scaling to 64 cpus on a sandard distribution. Where did I say more than 8 cpus? You are confusing unixconsole and Robert’s thread with mine,
Now pay attention and concentrate. All info you pointed to says that linux cannot scale to a 64 way configuration on a standard distro. Go back and show one statment I made that says I was arguing anything other than that.
All info you pointed to says that linux cannot scale to a 64 way configuration on a standard distro. Go back and show one statment I made that says I was arguing anything other than that.
ROFL. Seriously, I hope someone else is still reading this thread. I haven only said I was talking about Linux scaling past 8 CPUs about 50 times. If I was trying to show Linux scaling to 64 procs, why would I be pointing out 16 and 32 CPU systems? Because I wasn’t. Go back to being a fanboy, it’ll be easier for you.
Oh, btw, you need to learn more about your hardware. Every real RAID adapter has massive amounts of firmware on them. Every modern ethernet adapter has the firmware to handle the wire level portion of TCP. I could go on, but you stick to your SCSI adaptors or whatever it is you do. Obviously you know everything.
And when did I say Google has an SMP system? They use a massive cluster. I know exactly how it works. My comment was that Google also only runs 1 process on their cluster nodes. That process is the Google search application. They don’t boot up to a full distro, it’s a waste for them.
You fanboys are hysterical, I’m signing off, I wasted enough time pointing how rediculous you all are.
If I was trying to show Linux scaling to 64 procs, why would I be pointing out 16 and 32 CPU systems?
I was trying to figure that out myself. You are beyond help aren’t you. I proved you wrong and you change the subject. You actually are a disgrace to linux fanboys, most of the ones I debated so far had much more to offer.
Oh, btw, you need to learn more about your hardware. Every real RAID adapter has massive amounts of firmware on them. Every modern ethernet adapter has the firmware to handle the wire level portion of TCP. I could go on, but you stick to your SCSI adaptors or whatever it is you do. Obviously you know everything.
You never mentioned RAID. Show me the example of the ethernet with firware for wirelevel portion of TCP? Do you even know what TCP is? Why isn’t the wirelevel protocol IP? or ethernet frames? Do you know the OSI seven layer model and where TCP fits in it?
The problem is you don’t know squat but pretend you do.
And when did I say Google has an SMP system? They use a massive cluster. I know exactly how it works. My comment was that Google also only runs 1 process on their cluster nodes. That process is the Google search application. They don’t boot up to a full distro, it’s a waste for them.
Point me to the link that says so. Put up or shutup.
You fanboys are hysterical, I’m signing off, I wasted enough time pointing how rediculous you all are.
Good, Good riddance is what I’d say. Either you stop your boyish attitude and put some real facts, details and a real discussion or leave.
I think I have figured out Chris, he isn’t about presenting a coherent argument, he isn’t about presenting facts, he’s about being right and everybody else being wrong. So how about all of the “Solaris fanboys” start something new on OSNews and ignore the Linux trolls. We can relegate them to “BSD is dying” trolls on Slashdot.
Good Idea. I think ignoring them is far more productive than arguing with them. I’ll try and ignore them to the best of my abilities.