On 26 August 2011 13:30, Tudor Girba tudor@tudorgirba.com wrote:
Hi,
Since a couple of weeks I am working with the Cog VM built on Jenkins: https://ci.lille.inria.fr/pharo/view/Cog/job/Cog-Mac-Cocoa/
Here is a little feedback. It worked quite well, the only problem is that Monticello is really slow in the Moose image. This means that even opening the Monticello Browser takes several seconds.
I thought that this might be induced by something in the Moose image, but if I try with the Elliot's VM (2489), it works blazingly fast.
Thanks for the information, Tudor. Apparently the difference between Eliot's VM and Cog-Mac-Cocoa that they are using different frameworks (Carbon vs Cocoa) and i don't know too much details about the difference there. John or Esteban may have better insights how to fix that.
Cheers, Doru
-- www.tudorgirba.com
"The coherence of a trip is given by the clearness of the goal."
Tudor, can you give us some figures. Try to bench the code , like:
[ Gofer new blah blah load ] timeToRun
running in different VMs with same image. because the code which serving sockets are identical , so we should know for sure that there are some slowdowns or not, or maybe you was lucky enough to connect to the squeaksource when it was processing someone's uploads , or variance in network latency etc.
Hi,
I did not forget about this, but I did not have enough time yet.
Cheers, Doru
On 26 Aug 2011, at 14:40, Igor Stasenko wrote:
On 26 August 2011 13:30, Tudor Girba tudor@tudorgirba.com wrote:
Hi,
Since a couple of weeks I am working with the Cog VM built on Jenkins: https://ci.lille.inria.fr/pharo/view/Cog/job/Cog-Mac-Cocoa/
Here is a little feedback. It worked quite well, the only problem is that Monticello is really slow in the Moose image. This means that even opening the Monticello Browser takes several seconds.
I thought that this might be induced by something in the Moose image, but if I try with the Elliot's VM (2489), it works blazingly fast.
Thanks for the information, Tudor. Apparently the difference between Eliot's VM and Cog-Mac-Cocoa that they are using different frameworks (Carbon vs Cocoa) and i don't know too much details about the difference there. John or Esteban may have better insights how to fix that.
Cheers, Doru
-- www.tudorgirba.com
"The coherence of a trip is given by the clearness of the goal."
-- Best regards, Igor Stasenko AKA sig.
-- www.tudorgirba.com
"Yesterday is a fact. Tomorrow is a possibility. Today is a challenge."
Hi,
I looked a bit into the slowness of the OS X Jenkins Cog VM with Monticello. The problem exists and it is reproducible.
I performed the following experiments using the following script:
Time millisecondsToRun: [ 1 to: 100 do: [:x | | window | window := MCWorkingCopyBrowser new show. window delete ] ]
I ran it on the same image using the following vms (on Mac OS X Lion): - Jenkins VM from 31.08.2011 - Cog VM r2489
I ran the scenario twice with two different setups:
1. I placed the image in an empty folder:
Cog VM 2389: 4808 ms Jenkins VM: 7345 ms
2. I places the image in a folder that already had a large package-cache with 7400 mcz files: Cog VM: 4867 ms Jenkins VM: 50754 ms
So: - even if there is no package cache, the Jenkins VM seems to perform 1.5 times slower - when there is a package-cache, it becomes hardly unusable.
Cheers, Doru
On 8 Sep 2011, at 12:10, Tudor Girba wrote:
Hi,
I did not forget about this, but I did not have enough time yet.
Cheers, Doru
On 26 Aug 2011, at 14:40, Igor Stasenko wrote:
On 26 August 2011 13:30, Tudor Girba tudor@tudorgirba.com wrote:
Hi,
Since a couple of weeks I am working with the Cog VM built on Jenkins: https://ci.lille.inria.fr/pharo/view/Cog/job/Cog-Mac-Cocoa/
Here is a little feedback. It worked quite well, the only problem is that Monticello is really slow in the Moose image. This means that even opening the Monticello Browser takes several seconds.
I thought that this might be induced by something in the Moose image, but if I try with the Elliot's VM (2489), it works blazingly fast.
Thanks for the information, Tudor. Apparently the difference between Eliot's VM and Cog-Mac-Cocoa that they are using different frameworks (Carbon vs Cocoa) and i don't know too much details about the difference there. John or Esteban may have better insights how to fix that.
Cheers, Doru
-- www.tudorgirba.com
"The coherence of a trip is given by the clearness of the goal."
-- Best regards, Igor Stasenko AKA sig.
-- www.tudorgirba.com
"Yesterday is a fact. Tomorrow is a possibility. Today is a challenge."
-- www.tudorgirba.com
"Sometimes the best solution is not the best solution."
this is a job for the VM profiler. I expect it will zero in on the cause PDQ.
On Thu, Sep 22, 2011 at 10:32 AM, Tudor Girba tudor@tudorgirba.com wrote:
Hi,
I looked a bit into the slowness of the OS X Jenkins Cog VM with Monticello. The problem exists and it is reproducible.
I performed the following experiments using the following script:
Time millisecondsToRun: [ 1 to: 100 do: [:x | | window | window := MCWorkingCopyBrowser new show. window delete ] ]
I ran it on the same image using the following vms (on Mac OS X Lion):
- Jenkins VM from 31.08.2011
- Cog VM r2489
I ran the scenario twice with two different setups:
- I placed the image in an empty folder:
Cog VM 2389: 4808 ms Jenkins VM: 7345 ms
- I places the image in a folder that already had a large package-cache
with 7400 mcz files: Cog VM: 4867 ms Jenkins VM: 50754 ms
So:
- even if there is no package cache, the Jenkins VM seems to perform 1.5
times slower
- when there is a package-cache, it becomes hardly unusable.
Cheers, Doru
On 8 Sep 2011, at 12:10, Tudor Girba wrote:
Hi,
I did not forget about this, but I did not have enough time yet.
Cheers, Doru
On 26 Aug 2011, at 14:40, Igor Stasenko wrote:
On 26 August 2011 13:30, Tudor Girba tudor@tudorgirba.com wrote:
Hi,
Since a couple of weeks I am working with the Cog VM built on Jenkins: https://ci.lille.inria.fr/pharo/view/Cog/job/Cog-Mac-Cocoa/
Here is a little feedback. It worked quite well, the only problem is
that Monticello is really slow in the Moose image. This means that even opening the Monticello Browser takes several seconds.
I thought that this might be induced by something in the Moose image,
but if I try with the Elliot's VM (2489), it works blazingly fast.
Thanks for the information, Tudor. Apparently the difference between Eliot's VM and Cog-Mac-Cocoa that they are using different frameworks (Carbon vs Cocoa) and i don't know too much details about the difference there. John or Esteban may have better insights how to fix that.
Cheers, Doru
-- www.tudorgirba.com
"The coherence of a trip is given by the clearness of the goal."
-- Best regards, Igor Stasenko AKA sig.
-- www.tudorgirba.com
"Yesterday is a fact. Tomorrow is a possibility. Today is a challenge."
-- www.tudorgirba.com
"Sometimes the best solution is not the best solution."
On 22 September 2011 19:32, Tudor Girba tudor@tudorgirba.com wrote:
Hi,
I looked a bit into the slowness of the OS X Jenkins Cog VM with Monticello. The problem exists and it is reproducible.
I performed the following experiments using the following script:
Time millisecondsToRun: [ 1 to: 100 do: [:x | | window | window := MCWorkingCopyBrowser new show. window delete ] ]
I ran it on the same image using the following vms (on Mac OS X Lion):
- Jenkins VM from 31.08.2011
- Cog VM r2489
I ran the scenario twice with two different setups:
- I placed the image in an empty folder:
Cog VM 2389: 4808 ms Jenkins VM: 7345 ms
- I places the image in a folder that already had a large package-cache with 7400 mcz files:
Cog VM: 4867 ms Jenkins VM: 50754 ms
So:
- even if there is no package cache, the Jenkins VM seems to perform 1.5 times slower
- when there is a package-cache, it becomes hardly unusable.
Thanks, Tudor for sticking with it :)
so, your measurements telling that its actually could be related more to filesystem interaction(s) rather than to network.
we need to ask an experts (Esteban and John) , what are difference between cocoa and carbon versions in file handling, that may cause such big slowdown.
Cheers, Doru
On 8 Sep 2011, at 12:10, Tudor Girba wrote:
Hi,
I did not forget about this, but I did not have enough time yet.
Cheers, Doru
On 26 Aug 2011, at 14:40, Igor Stasenko wrote:
On 26 August 2011 13:30, Tudor Girba tudor@tudorgirba.com wrote:
Hi,
Since a couple of weeks I am working with the Cog VM built on Jenkins: https://ci.lille.inria.fr/pharo/view/Cog/job/Cog-Mac-Cocoa/
Here is a little feedback. It worked quite well, the only problem is that Monticello is really slow in the Moose image. This means that even opening the Monticello Browser takes several seconds.
I thought that this might be induced by something in the Moose image, but if I try with the Elliot's VM (2489), it works blazingly fast.
Thanks for the information, Tudor. Apparently the difference between Eliot's VM and Cog-Mac-Cocoa that they are using different frameworks (Carbon vs Cocoa) and i don't know too much details about the difference there. John or Esteban may have better insights how to fix that.
Cheers, Doru
-- www.tudorgirba.com
"The coherence of a trip is given by the clearness of the goal."
-- Best regards, Igor Stasenko AKA sig.
-- www.tudorgirba.com
"Yesterday is a fact. Tomorrow is a possibility. Today is a challenge."
-- www.tudorgirba.com
"Sometimes the best solution is not the best solution."
Hi,
On 22 Sep 2011, at 19:59, Igor Stasenko wrote:
On 22 September 2011 19:32, Tudor Girba tudor@tudorgirba.com wrote:
Hi,
I looked a bit into the slowness of the OS X Jenkins Cog VM with Monticello. The problem exists and it is reproducible.
I performed the following experiments using the following script:
Time millisecondsToRun: [ 1 to: 100 do: [:x | | window | window := MCWorkingCopyBrowser new show. window delete ] ]
I ran it on the same image using the following vms (on Mac OS X Lion):
- Jenkins VM from 31.08.2011
- Cog VM r2489
I ran the scenario twice with two different setups:
- I placed the image in an empty folder:
Cog VM 2389: 4808 ms Jenkins VM: 7345 ms
- I places the image in a folder that already had a large package-cache with 7400 mcz files:
Cog VM: 4867 ms Jenkins VM: 50754 ms
So:
- even if there is no package cache, the Jenkins VM seems to perform 1.5 times slower
- when there is a package-cache, it becomes hardly unusable.
Thanks, Tudor for sticking with it :)
so, your measurements telling that its actually could be related more to filesystem interaction(s) rather than to network.
This is what I understood as well.
we need to ask an experts (Esteban and John) , what are difference between cocoa and carbon versions in file handling, that may cause such big slowdown.
Cheers, Doru
On 8 Sep 2011, at 12:10, Tudor Girba wrote:
Hi,
I did not forget about this, but I did not have enough time yet.
Cheers, Doru
On 26 Aug 2011, at 14:40, Igor Stasenko wrote:
On 26 August 2011 13:30, Tudor Girba tudor@tudorgirba.com wrote:
Hi,
Since a couple of weeks I am working with the Cog VM built on Jenkins: https://ci.lille.inria.fr/pharo/view/Cog/job/Cog-Mac-Cocoa/
Here is a little feedback. It worked quite well, the only problem is that Monticello is really slow in the Moose image. This means that even opening the Monticello Browser takes several seconds.
I thought that this might be induced by something in the Moose image, but if I try with the Elliot's VM (2489), it works blazingly fast.
Thanks for the information, Tudor. Apparently the difference between Eliot's VM and Cog-Mac-Cocoa that they are using different frameworks (Carbon vs Cocoa) and i don't know too much details about the difference there. John or Esteban may have better insights how to fix that.
Cheers, Doru
-- www.tudorgirba.com
"The coherence of a trip is given by the clearness of the goal."
-- Best regards, Igor Stasenko AKA sig.
-- www.tudorgirba.com
"Yesterday is a fact. Tomorrow is a possibility. Today is a challenge."
-- www.tudorgirba.com
"Sometimes the best solution is not the best solution."
-- Best regards, Igor Stasenko.
-- www.tudorgirba.com
"One cannot do more than one can do."
Check to see if your VM has ZipPlugin (Smalltalk listBuiltinModules). This has a big effect on Monticello performance.
Dave
On Thu, Sep 22, 2011 at 07:32:17PM +0200, Tudor Girba wrote:
Hi,
I looked a bit into the slowness of the OS X Jenkins Cog VM with Monticello. The problem exists and it is reproducible.
I performed the following experiments using the following script:
Time millisecondsToRun: [ 1 to: 100 do: [:x | | window | window := MCWorkingCopyBrowser new show. window delete ] ]
I ran it on the same image using the following vms (on Mac OS X Lion):
- Jenkins VM from 31.08.2011
- Cog VM r2489
I ran the scenario twice with two different setups:
- I placed the image in an empty folder:
Cog VM 2389: 4808 ms Jenkins VM: 7345 ms
- I places the image in a folder that already had a large package-cache with 7400 mcz files:
Cog VM: 4867 ms Jenkins VM: 50754 ms
So:
- even if there is no package cache, the Jenkins VM seems to perform 1.5 times slower
- when there is a package-cache, it becomes hardly unusable.
Cheers, Doru
On 8 Sep 2011, at 12:10, Tudor Girba wrote:
Hi,
I did not forget about this, but I did not have enough time yet.
Cheers, Doru
On 26 Aug 2011, at 14:40, Igor Stasenko wrote:
On 26 August 2011 13:30, Tudor Girba tudor@tudorgirba.com wrote:
Hi,
Since a couple of weeks I am working with the Cog VM built on Jenkins: https://ci.lille.inria.fr/pharo/view/Cog/job/Cog-Mac-Cocoa/
Here is a little feedback. It worked quite well, the only problem is that Monticello is really slow in the Moose image. This means that even opening the Monticello Browser takes several seconds.
I thought that this might be induced by something in the Moose image, but if I try with the Elliot's VM (2489), it works blazingly fast.
Thanks for the information, Tudor. Apparently the difference between Eliot's VM and Cog-Mac-Cocoa that they are using different frameworks (Carbon vs Cocoa) and i don't know too much details about the difference there. John or Esteban may have better insights how to fix that.
Cheers, Doru
-- www.tudorgirba.com
"The coherence of a trip is given by the clearness of the goal."
-- Best regards, Igor Stasenko AKA sig.
-- www.tudorgirba.com
"Yesterday is a fact. Tomorrow is a possibility. Today is a challenge."
-- www.tudorgirba.com
"Sometimes the best solution is not the best solution."
On Thu, Sep 22, 2011 at 9:57 PM, David T. Lewis lewis@mail.msen.com wrote:
Check to see if your VM has ZipPlugin (Smalltalk listBuiltinModules).
Good point. I see it. Doru, do you ?
This has a big effect on Monticello performance.
Dave
On Thu, Sep 22, 2011 at 07:32:17PM +0200, Tudor Girba wrote:
Hi,
I looked a bit into the slowness of the OS X Jenkins Cog VM with
Monticello. The problem exists and it is reproducible.
I performed the following experiments using the following script:
Time millisecondsToRun: [ 1 to: 100 do: [:x | | window | window := MCWorkingCopyBrowser new show. window delete ] ]
I ran it on the same image using the following vms (on Mac OS X Lion):
- Jenkins VM from 31.08.2011
- Cog VM r2489
I ran the scenario twice with two different setups:
- I placed the image in an empty folder:
Cog VM 2389: 4808 ms Jenkins VM: 7345 ms
- I places the image in a folder that already had a large package-cache
with 7400 mcz files:
Cog VM: 4867 ms Jenkins VM: 50754 ms
So:
- even if there is no package cache, the Jenkins VM seems to perform 1.5
times slower
- when there is a package-cache, it becomes hardly unusable.
Cheers, Doru
On 8 Sep 2011, at 12:10, Tudor Girba wrote:
Hi,
I did not forget about this, but I did not have enough time yet.
Cheers, Doru
On 26 Aug 2011, at 14:40, Igor Stasenko wrote:
On 26 August 2011 13:30, Tudor Girba tudor@tudorgirba.com wrote:
Hi,
Since a couple of weeks I am working with the Cog VM built on
Jenkins:
https://ci.lille.inria.fr/pharo/view/Cog/job/Cog-Mac-Cocoa/
Here is a little feedback. It worked quite well, the only problem is
that Monticello is really slow in the Moose image. This means that even opening the Monticello Browser takes several seconds.
I thought that this might be induced by something in the Moose image,
but if I try with the Elliot's VM (2489), it works blazingly fast.
Thanks for the information, Tudor. Apparently the difference between Eliot's VM and Cog-Mac-Cocoa that they are using different frameworks (Carbon vs Cocoa) and i don't know too much details about the difference there. John or Esteban may have better insights how to fix that.
Cheers, Doru
-- www.tudorgirba.com
"The coherence of a trip is given by the clearness of the goal."
-- Best regards, Igor Stasenko AKA sig.
-- www.tudorgirba.com
"Yesterday is a fact. Tomorrow is a possibility. Today is a challenge."
-- www.tudorgirba.com
"Sometimes the best solution is not the best solution."
vm-dev@lists.squeakfoundation.org
-
David T. Lewis -
Eliot Miranda -
Igor Stasenko -
Mariano Martinez Peck -
Tudor Girba