R help - Jul 2012 - Two R sessions on multicore computer seem to inhibit each other ?

Dear R-helpers,

I am puzzled by the following observation: 
On my home dual core Windows desktop computer, I am used to running two R
sessions in parallel. These do very well in using the full CPU of the
computer (half each) and don't seem to slow each other down.

Today I have started some large computation effort in one of our university
labs, and I intended to go for two R sessions per computer as well. These
computers are all identical and have Intel Core i5, 3.3 GHz processors with
four cores, and they are used for my computations and nothing else at the
moment. When I run one R session (32 bit) on them, the Windows task manager
shows CPU usage at about 25%. If I add another R session that should also
use as much CPU as it can get, the windows task manager showed overall CPU
usage of the two sessions below 25%. 
This made me very suspicious, and I decided to use separate computers for
all tasks (glad I did) and started a small investigation into what is going
on here: one extra computer runs two R sessions that do the exact same
calculations done on two other computers. I am surprised to see the enormous
speed advantage of doing calculations on separate computers: a calculation
that took 15 minutes on a computer that ran a single R session took 156
minutes in an R session on a computer with two R sessions running; the other
example: 8 minutes to 71 minutes. 

Does anyone have any idea why running two R sessions simultaneously slows
everything down so much? 

Thanks and regards,
Ulrike

--
View this message in context:
http://r.789695.n4.nabble.com/Two-R-sessions-on-multicore-computer-seem-to-inhibit-each-other-tp4636336.html
Sent from the R help mailing list archive at Nabble.com.

P.S.: I should have mentioned: The operating system is Windows XP.

--
View this message in context:
http://r.789695.n4.nabble.com/Two-R-sessions-on-multicore-computer-seem-to-inhibit-each-other-tp4636336p4636355.html
Sent from the R help mailing list archive at Nabble.com.

My first guess would be there's not enough memory on the machine to support
two of the tasks you're running, so swapping ensues.

--Todd
-- 
Why is it so hard to remember the spelling of "mnemonic"?


-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org]
On Behalf Of Ulrike Gr?mping
Sent: Thursday, July 12, 2012 10:55 AM
To: r-help at r-project.org
Subject: [R] Two R sessions on multicore computer seem to inhibit each other ?

Dear R-helpers,

I am puzzled by the following observation: 
On my home dual core Windows desktop computer, I am used to running two R
sessions in parallel. These do very well in using the full CPU of the
computer (half each) and don't seem to slow each other down.

Today I have started some large computation effort in one of our university
labs, and I intended to go for two R sessions per computer as well. These
computers are all identical and have Intel Core i5, 3.3 GHz processors with
four cores, and they are used for my computations and nothing else at the
moment. When I run one R session (32 bit) on them, the Windows task manager
shows CPU usage at about 25%. If I add another R session that should also
use as much CPU as it can get, the windows task manager showed overall CPU
usage of the two sessions below 25%. 
This made me very suspicious, and I decided to use separate computers for
all tasks (glad I did) and started a small investigation into what is going
on here: one extra computer runs two R sessions that do the exact same
calculations done on two other computers. I am surprised to see the enormous
speed advantage of doing calculations on separate computers: a calculation
that took 15 minutes on a computer that ran a single R session took 156
minutes in an R session on a computer with two R sessions running; the other
example: 8 minutes to 71 minutes. 

Does anyone have any idea why running two R sessions simultaneously slows
everything down so much? 

Thanks and regards,
Ulrike

--
View this message in context:
http://r.789695.n4.nabble.com/Two-R-sessions-on-multicore-computer-seem-to-inhibit-each-other-tp4636336.html
Sent from the R help mailing list archive at Nabble.com.

______________________________________________
R-help at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

what would be nice is to understand the amount of cpu time vs. the elapsed time
consumed.  you might be paging if you don't have sufficient real memory.  I
would venture a guess that the cpu time is very similar and the differnce is
elapsed time due to some contention like memory.

Sent from my iPad

On Jul 12, 2012, at 13:54, Ulrike Gr?mping <groemping at bht-berlin.de>
wrote:
> Dear R-helpers,
> 
> I am puzzled by the following observation: 
> On my home dual core Windows desktop computer, I am used to running two R
> sessions in parallel. These do very well in using the full CPU of the
> computer (half each) and don't seem to slow each other down.
> 
> Today I have started some large computation effort in one of our university
> labs, and I intended to go for two R sessions per computer as well. These
> computers are all identical and have Intel Core i5, 3.3 GHz processors with
> four cores, and they are used for my computations and nothing else at the
> moment. When I run one R session (32 bit) on them, the Windows task manager
> shows CPU usage at about 25%. If I add another R session that should also
> use as much CPU as it can get, the windows task manager showed overall CPU
> usage of the two sessions below 25%. 
> This made me very suspicious, and I decided to use separate computers for
> all tasks (glad I did) and started a small investigation into what is going
> on here: one extra computer runs two R sessions that do the exact same
> calculations done on two other computers. I am surprised to see the
enormous
> speed advantage of doing calculations on separate computers: a calculation
> that took 15 minutes on a computer that ran a single R session took 156
> minutes in an R session on a computer with two R sessions running; the
other
> example: 8 minutes to 71 minutes. 
> 
> Does anyone have any idea why running two R sessions simultaneously slows
> everything down so much? 
> 
> Thanks and regards,
> Ulrike
> 
> --
> View this message in context:
http://r.789695.n4.nabble.com/Two-R-sessions-on-multicore-computer-seem-to-inhibit-each-other-tp4636336.html
> Sent from the R help mailing list archive at Nabble.com.
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

Thanks to both of you, you are probably right that memory is the limiting
factor.  I have no knowledge about the available memory on the lab machines,
but I will find out and make sure that this is the explanation.

Best,
Ulrike

--
View this message in context:
http://r.789695.n4.nabble.com/Two-R-sessions-on-multicore-computer-seem-to-inhibit-each-other-tp4636336p4636396.html
Sent from the R help mailing list archive at Nabble.com.