I hope this is the right forum to post this question.
We have SLES 11 running on a DELL server.
My IT department ran an engineering analysis software, Abaqus, on a trial version of SLED and a trial version of Windows Server 2012 on a workstation, a Dell T5500, and found that the analysis completed in Windows 12% quicker than it did in SLED.
This surprised everyone as we expected Windows to lag behind Linux. Abaqus is certified to run on SLES 11. My question is, will applications built for SLES run the same in SLED?
Is SLED as robust as SLES or is it just a hacked down version of SLES?
I know its not fair to compare SLED against Windows Server. But both operating systems were running of a mid range workstation, so I would have expected comparable run times.
I hope this is the right forum to post this question.
We have SLES 11 running on a DELL server.
My IT department ran an engineering analysis software, Abaqus, on a trial version of SLED and a trial version of Windows Server 2012 on a workstation, a Dell T5500, and found that the analysis completed in Windows 12% quicker than it did in SLED.
This surprised everyone as we expected Windows to lag behind Linux. Abaqus is certified to run on SLES 11. My question is, will applications built for SLES run the same in SLED?
Is SLED as robust as SLES or is it just a hacked down version of SLES?
I know its not fair to compare SLED against Windows Server. But both operating systems were running of a mid range workstation, so I would have expected comparable run times.
Thanks.[/QUOTE]
Hi
Different kernel optimizations for a start, then you get into other performance optimizations (eg I/O, memory), so yes I would expect a difference between SLES and SLED.
Why not grab an evaluation of SLES 11 and give that a try.
[QUOTE=intecperict;20944]Thanks Malcolm, so I was not wrong in assuming there is a difference between SLED and SLES.
Thanks again.[/QUOTE]
Hi
Just did a quick to compare kernels via comparing the kernel configs, and there is no difference for the current 3.0.101-0.21-default so you would need to compare the sysctl outputs, did a quick check here and there are numerous differences and I’m sure there are other places… esp. hardware wise done during the install.
But I’m sure SLED could be tweaked to a certain degree to make it similar.
It’s only a fair comparison of how the Linux version of Abaqus performs vs the Windows version on the same hardware. That could depend on factors such as how much effort it’s authors have put in to performance optimisations in each version. The title you used for the thread is “Is SLED Slower than Windows Server 2012?” which is a very broad question. What you’re actually asking is “Is SLED slow than Windows Server 2012 for my particular use case?” For some purposes it might be slower. For some purposes it might be quicker.
Mike, you’re absolutely right in saying that. I should have been more specific when asking my question.
After my IT in the US checked his configuration, it turns out that he was using SLES all the time. Which makes me think now that Abaqus runs faster on Windows than on SLES, which is a shame as that would mean they would eventually wipe clean the SLES server and have Windows Server running on it.
I never thought I’d see the day when software developers start fine tuning their apps for Windows.
I am sorry for not properly addressing the subject of this post. But I really appreciated all of your advices.
[QUOTE=intecperict;21018]Mike, you’re absolutely right in saying that. I should have been more specific when asking my question.
After my IT in the US checked his configuration, it turns out that he was using SLES all the time. Which makes me think now that Abaqus runs faster on Windows than on SLES, which is a shame as that would mean they would eventually wipe clean the SLES server and have Windows Server running on it.
I never thought I’d see the day when software developers start fine tuning their apps for Windows.
I am sorry for not properly addressing the subject of this post. But I really appreciated all of your advices.[/QUOTE]
Hi
So this is an engineering program? If so are you using Nvida gpu’s, these drivers are separate installs on SLE as are non-oss in nature.
Would be interesting to see the hardware config and software config on the windows server
[QUOTE=malcolmlewis;21037]Hi
So this is an engineering program? If so are you using Nvida gpu’s, these drivers are separate installs on SLE as are non-oss in nature.
Would be interesting to see the hardware config and software config on the windows server ;)[/QUOTE]
Malcolm, yes, its a finite element analysis software and its very disk and RAM intensive, so I/O is always an issue. Some of our current runs comsume around 45GB of RAM per job. That’s why we opted for Linux as we didnt think Windows could cope with such a high demand. The machine has 256GB of RAM at the moment.
I dont have the windows server 2012 software config as it was done by my IT colleague in the US but I would assume just a standard Windows Server 2012 installation. Both Windows Server and SLES were installed right out of the box without any modification or optimization.
My last question and I know this is the wrong forum for it since it should be in the SLES forum but before I post this question and explain everything again to the SLES forum, I would like to know if SLES can be tweaked for better performance or is out-of-the-box installation already and optimized solution?
performance optimization is always the job of configuring the right system for the right task. SLES (as well as SLED) is “optimized” for some typical workload expected to run on the systems, so if you run some non-typical program, there’s room left for improvement. Especially when you only run one specific (type of) task, as you can fully focus on that one, rather then creating a system that has to compromise between different types of tasks.
