Linked by Thom Holwerda on Wed 18th Mar 2009 11:48 UTC, submitted by PLan
In the News In a move that would certainly shake up the computer industry quite a bit, IBM is reportedly in talks with Sun Microsystems about the possibility of IBM acquiring Sun. Sun is going through hard times at the moment, and has been actively looking for someone to be acquired by.
Thread beginning with comment 354380
To view parent comment, click here.
To read all comments associated with this story, please click here.
RE[6]: Will OpenSolaris survive?
by segedunum on Sun 22nd Mar 2009 01:04 UTC in reply to "RE[5]: Will OpenSolaris survive?"
segedunum
Member since:
2005-07-06

I think I even overestimated Linux. Maybe 32 cores max. But enlighten me, show me a link to x86 server with more than 32 (okay, maybe 48) cores.

I'm not entirely sure why you're talking about 'cores', but if it's actually SMP systems you're talking about (it can't be anything else) then I think you've been living in the dark ages somewhere. Either that or you believe Sun's marketing literature.

Trust me, it's been done and people are running stuff with an awful lot more than 32 'cores' in them. Have been for years. To post links would be fruitless and would merely insult your ability to use Google. If that's beyond you however then SGI have been doing the really big stuff, and that's with kit with >= 1024 CPUs in them. They're enormous:

http://marc.info/?l=linux-kernel&m=117580267612045&w=2

Certainly at the time that that thread was posted Solaris had never ran on anything like a 1024 CPU system. To top it off, Solaris doesn't do RCU locking either:

http://lse.sourceforge.net/locking/rcupdate.html

Linux not scalable? Solaris more scalable? More 'solid'? I Wouldn't call it FUD. It's just outright desperation, but hey, that's why Sun are in trouble.

I could tell you stories from experience how "unbrekable" Oracle Linux is on 8 core AMD server.

Dunno. There's umpteen people doing it, so maybe it's just you? I could tell you many stories about Solaris's exceptionally buggy drivers, many of them IDE drivers that Sun is only now finding out about by using ZFS as a debugging tool, how difficult it is to get software actually installed on Solaris and how it has had to be explained to Sun's consultants that no, we aren't going to make SPARC faster by recompiling our open source software in Forte.

However, if you chose Oracle Linux then you probably don't know what you're doing anyway so it's all academic really.

Edited 2009-03-22 01:14 UTC

Reply Parent Score: 2

dvzt Member since:
2008-10-23

I'm not entirely sure why you're talking about 'cores'


Because there is difference between a single core-cpu and multi-core one. OS sees multiple virtual processors.

but if it's actually SMP systems you're talking about (it can't be anything else)


I'm talking about NUMA.

Trust me, it's been done and people are running stuff with an awful lot more than 32 'cores' in them. Have been for years. To post links would be fruitless and would merely insult your ability to use Google. If that's beyond you however then SGI have been doing the really big stuff, and that's with kit with >= 1024 CPUs in them. They're enormous:

http://marc.info/?l=linux-kernel&m=117580267612045&w=2

Certainly at the time that that thread was posted Solaris had never ran on anything like a 1024 CPU system.


A server and supercomputer are two very diferent things. Supercomputers are usually used for simulations and data processing, whereas servers run tradional stuff like applications and databases. Server's OS usually has quite difficult task of managing hundreds of processes competing for same resources like disk, memory and network, but a supercomputer just needs to do maths and that's it. One process on each cpu is a very likely setup. If an OS runs on a supercomputer it proves absolutely nothing. People put Linux on supercomputers because it's free and easily customizable.

To top it off, Solaris doesn't do RCU locking either:

http://lse.sourceforge.net/locking/rcupdate.html


Interresting link, but since I never heared of it, it's probably not so huge.

Linux not scalable? Solaris more scalable? More 'solid'? I Wouldn't call it FUD. It's just outright desperation,


Not FUD, not desperation, but a fact. Now it's time for me to ask you, where have you been living? Linux is just a low-cost OS for a low cost, low end servers (x86). Take Sun Fire M9000, that's 64 4-core, 8 thread cpus, i.e. OS sees 512 virtual cpus. Corporations run their mission critical stuff like Oracle databases on it. You don't see anything like that with Linux, ever. Besides, for a server availability is usually more important, because if your applications don't work, employees can't do their jobs and you loose money. If your supercomputer goes down, you maybe won't finish your scientific calculations this week - no big deal, you'll get it next week.
Seriously, where have you been hiding? All the long-time sysadmins don't even take Linux seriously.

Linux solid? You call an OS with no stable DDI solid? Take VxVM for example. It has at least dozen different kernel module packages for different kernel versions. That's a laugh! No other OS (including Windows) needs that. You call OS with multiple incompatible versions solid? And I could go on...

but hey, that's why Sun are in trouble.


The cause of their problem lies with bad marketing, but I'm not going to discuss that.

"I could tell you stories from experience how "unbrekable" Oracle Linux is on 8 core AMD server.

Dunno. There's umpteen people doing it, so maybe it's just you?
"

It's not me, it's people from oracle who were doing the implementation as part of a normal paid project. I can't think of anyone more competent.

I could tell you many stories about Solaris's exceptionally buggy drivers,


Why not? Every piece of software encouters bugs.

many of them IDE drivers that Sun is only now finding out about by using ZFS as a debugging tool,


You have IDE drives in your server? Wow.

how difficult it is to get software actually installed on Solaris


How? I never had problems, so maybe it's just you.

However, if you chose Oracle Linux then you probably don't know what you're doing anyway so it's all academic really.


I'm not the decision maker here. It's true, our IT architects failed this time.

Reply Parent Score: 1

Kebabbert Member since:
2007-07-27

"Seriously, where have you [Segedunum] been hiding? All the long-time sysadmins don't even take Linux seriously."

Look, this self contradictory, very selective, Segedunum also states that VirtualBox is equally stable, or more stable, than VMware. If he can state that weird thing, he can as well state that Linux is more stable than Solaris and AIX, or "the moon is made of cheese".

Reply Parent Score: 2

segedunum Member since:
2005-07-06

Because there is difference between a single core-cpu and multi-core one. OS sees multiple virtual processors.

Yes. They're called SMP systems regardless of whether they use cores in one processor or multiple processors. The distinction makes no difference to the OS or its scalability.

I'm talking about NUMA.

No you're not. You don't know what you're talking about.

A server and supercomputer are two very diferent things.

No they're not. You wanted to try and throw a blanket statement over Linux's scalability and you've been shown to be wrong.

Interresting link, but since I never heared of it, it's probably not so huge.

Yes, because you don't know what you're talking about. You profess to cast doubt on Linux's scalibility where SMP is concerned without knowing you're talking about regarding SMP.

Linux is just a low-cost OS for a low cost, low end servers (x86).......Seriously, where have you been hiding? All the long-time sysadmins don't even take Linux seriously.

Sun has been saying that for ten years, and has resulted in this article about takeover rumours. It obviously hasn't worked because it isn't true. Deal with it.

Edited 2009-03-22 15:09 UTC

Reply Parent Score: 2

Kebabbert Member since:
2007-07-27

SEGEDUNUM

What is the problem with you?

I (and others) have told you several times that ZFS doesnt need huge amounts of GB to run. ZFS runs fine on 1GB RAM (Ive run ZFS on 1GB for over a year). And still, in every post about ZFS you tell people that ZFS needs huge amounts of RAM, several GB. And you tell people that ZFS end-to-end data integrity is not a big deal, even though it protects against silent corruption. The point of using ZFS is it's end-to-end data integrity. All the rest is just icing on the cake.

(Ive heard that FreeBSD implementation of ZFS takes much RAM. But that is not proof of ZFS needing much RAM. That is proof that FreeBSD implementation needs much RAM. Not ZFS. Easy to understand the distinction, if you have learned to draw correct conclusions)

We told you that several times, and you just dont listen. You dont get it. Or, choose not to understand. Very selective. Whenever you hear something bad about ZFS or a SUN product, it sticks. When you hear opposite, you just ignore it. Very selective.

Or, is it that you can not keep two contradicting facts in your mind, at the same time? You have to choose between one of them? And then you choose randomly? Or you keep the fact negative to ZFS and SUN? Your L2 cache only fit one piece of fact?





Regarding that SGI machine with Linux. I dont know how many times I have to tell you. I explained this for you recently, and I will have to explain again. I suspect.

Linux on SGI runs a modified custom made kernel. It is not stock Linux. I can modify MS-DOS and run on it several CPUs. BUT THAT IS NOT SCALABILITY, IT IS MODIFIABILITY! But Solaris kernel is the same that runs on machines from Intel Atom to huge servers. THAT is scalability. There are no different versions of Solaris kernel.

But well, I guess there was just a chance epsilon big that you got that. I promise, you will continue to state that Linux is scalable. It is not. If unmodified Linux would run on small devices up to huge servers and to large clusters, then Linux would be scalable. Otherwise, I can modify MS-DOS and call it scalable, using the same reasoning as you do. But noone would call MS-DOS scalable, would they? But, hey, "Linux is scalable", but MS-DOS is not???? Whats the difference? Clusters dont count as big iron.

There are some contradiction here. I see that, others see that. You dont see that.

Edited 2009-03-22 09:31 UTC

Reply Parent Score: 2

segedunum Member since:
2005-07-06

What is the problem with you?

What's wrong with you lot? Sun's the one wanting to be sold here. There is obviously sometjhing wrong and I'm afraid you just have a real mental issue over it.

I (and others) have told you several times that ZFS doesnt need huge amounts of GB to run. ZFS runs fine on 1GB RAM (Ive run ZFS on 1GB for over a year).

I'm afraid that has absolutely zilch to do with what I talked about in this thread, and has has been explained to you numpties on umpteen occasions about ZFS, just because you've ran it on your dinky little laptop for a year it proves jack. The BSD developers have certainly found that it will eat memory unbounded relative to the workload you throw at it. The end.

And still, in every post about ZFS.....

I never talked abiut ZFS anywhere, this thread was never about ZFS other than using it as a debugging tool for Solaris's buggy IDE drivers.

I feel for you when the inevitable happens and Sun is taken over. Obviously many people have a pretty big emotional attachment and it will hit them hard.

Reply Parent Score: 2