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 354087
To read all comments associated with this story, please click here.
Linux doesnt scale
by Kebabbert on Fri 20th Mar 2009 09:22 UTC
Kebabbert
Member since:
2007-07-27

Ive heard there are people claiming Linux scales well. Well, they are wrong. And maybe less intelligent to state such an argument, because they draw the wrong conclusions (lack of logic and math studies).

The thing is, Linux runs on big clusters, yes. But that doesnt means Linux scales well. In fact, Linux on big clusters is a modified Linux kernel. It is not stock Linux kernel. I could take C64 emulator and make it run on big clusters by redesigning the code, but I would be quite dumb if I claimed that C64 scales well.

Linux runs on wristwatches to large clusters, but it is a modified non standard Linux kernel. Of course you could modify C64 or Solaris or MS-DOS to do that as well. That is not a proof of scalability, it is a proof of Linux being modifiable.





When we talk about scalability, (llama): We're not talking about clusters. We're talking about single-system-image big iron, where _one_ kernel runs on a single machine with > 16 CPUs in a cache-coherent shared-memory system. The most cost-effective machines for cluster-building, in CPU power per dollar, are dual-socket quad core Intel Core2-based machines. i.e. 8 cores per node. That's great if you have a workload that has some coarse-grained parallelism, or is embarrassingly parallel, e.g. processing 100 separate data sets with single-thread processes that don't depend on each other. That's not so great if you have a lot of processes that need fine-grained access to the same shared resource. The canonical example here is a database server handling a database with a significant amount of write accesses. Otherwise you could just replicate it to a big cluster and spread the read load around. Locking for write access in a big cluster, even with low latency interconnects like infiniband, is still _way_ higher overhead than you'd get in e.g. a 4 or 8 socket quad-core machine. Even NUMA big iron is better suited for this than a cluster.

CLUSTERS DON'T COUNT AS BIG IRON. They're just a pile of normal machines. They do have their uses, though.






Linux scaling experts dispells the FUD of Linux scales bad. They spend lots of energy to explain that Linux scaling bad is only FUD from evil vendors, and that Linux scales very good.
http://searchenterpriselinux.techtarget.com/news/article/0,289142,s...
Lastly he says things as:

"In 1998, Linux would successfully run on one to two CPUs. Today, it can handle eight or more CPUs."

"With the 2.6 kernel, the vertical scaling will improve to 16-way."
Wow! 16 cpus! That is really really good! (NOT). Why does he talk about large Linux clusters with thousands of machines as a proof on Linux scales well, and at the same time declares that in several years from now, Linux will handle 16 cpus? Is 16 CPUs something to brag about?

"Question: Two years from now, where will Linux be, scalability-wise, in comparison to Windows and Unix?
O'Keefe: It will be at least comparable in most areas."
Wow. Bad Linux scaling is really FUD. I am convinced now. *sarcasm*

Morons.






Solaris scales from Intel Atom up to big iron with hundreds of CPUs worth millions of USD. With the EXACTLY the same install DVD. Now THAT is scalability. Modifiability is not scalability. Morons.
http://queue.acm.org/detail.cfm?id=1095419






Ive seen articles were Linux is faster than Solaris. Upon closer scrutiny, they compared an ancient 800MHz SPARC vs dual core Xeon 2.67GHz Linux. Another article stated their new Linux solution is faster than their old Solaris solution. Upon closer scrutiny, their Solaris solution were 800 old SPARC servers and their new Linux solution is 3000 new Linux dual Intel Xeon servers.

Morons.






In fact, several companies tries out Linux, and when their workload increases, they find out Linux doesnt cut it anymore. They switch to Solaris:
http://lethargy.org/~jesus/archives/77-Choosing-Solaris-10-over-Lin...

Linux sucks badly as a file server:
http://www.enterprisestorageforum.com/sans/features/article.php/374...

Linux Kernel developer Andrew Morton explains that Linux kernel is so full of bugs the quality is declining:
http://lwn.net/Articles/285088/

Only recently Linux has removed a barrier, it is now 250 times faster on 64 CPU systems. Wow. Linux was 250 times slower recently on 64 CPU systems. What other barriers are there? Linux have problems with scaling above 16 CPUs.



Trasz writes:
"The main disadvantage [Linux Kernel] of giant lock is that it eliminates the concurrency, thus decrease the performance on multiprocessor systems.
...
What I was talking about was synchronisation in the [Linux] kernel. What you're talking about above is synchronisation between userland threads. Two completely unrelated things. The fact that Linux uses spinlocks is one of the reasons that its performance drops noticeably under high load on many CPUs. Other operating systems use fully functional mutexes, along with interrupt threads.
...
This was about replacing so called 'semaphores' (actually, Linux' implementation of semaphores) with so called 'mutexes'. Spin locks are still the fundamental synchronisation mechanism."





For CDDL vs GPL, the problem with CDDL is that it allows CDDL on indvidual files. That is why Apple and BSD can lift in the ZFS source code files and use it. Whereas GPL demands that all files must be GPL, no mixing of licenses allowed. Some say that GPL is quite ego centric license. Everything revolves round GPL. No other licenses allowed. GPL is the TRUE license. Ego centric POW?






SEGEDUNUM,
As much as I am a SUN fanboy, even I know that VMware is for production and VirtualBox is not. VirtualBox is unstable - contrary to your claim. That is an ignorant claim and only FUD. As usual from your side. I would love if SUN's VB could compete with VMware in terms of stability, but that is unfortunately not true yet. It is obvious for someone who studied hight math: VB supports lots of platforms and OSes. VMware focuses on only a few. It is easier to get a few platforms stable, than all OSes on the whole market. Obvious. Except for your intellect.

Edited 2009-03-20 09:24 UTC

Reply Score: 1