[Pharo-project] OT: vm performance
Hello all, Here I am referring to a vm in the OS sense. Suppose you are forced to run code on a virtual machine running on a blade computer hidden behind locked doors. Access to the machine is limited to a remote desktop connection. You suspect that the machine is far too busy for its own good. Is there any way to test that? Is there any way to separate network latency from the box being slow, or perhaps just plain busy restoring a backup to fix up the most recent crash of some piece of hardware or who-knows-what software? Something that periodically runs a lightweight benchmark and logs the results comes to mind. Is there something that is commonly used for such measurments? I would rather not add too badly to machine's workload, but it gets sloooooowwwwwww at times :( Getting some data to objectify that statement could be useful. Bill
There was code supplied by both Andreas and Bert and corrected by Juan that addressed the issue of profiling a server image Subject: Re: [squeak-dev] Re: How to profile a server image? Date: March 4, 2009 5:07:52 AM PST Perhaps you have a seaside component that kicks off a MessageTally then it shows you the results to determine if things look correct. You'll need to follow the emails and see if any of the changes actually made it into Pharo, it would have been helpful if they were. You could use OSProcess to pull the results of the command 'uptime' so you can understand how busy it is. On 2009-09-28, at 4:37 PM, Schwab,Wilhelm K wrote:
Hello all,
Here I am referring to a vm in the OS sense. Suppose you are forced to run code on a virtual machine running on a blade computer hidden behind locked doors. Access to the machine is limited to a remote desktop connection. You suspect that the machine is far too busy for its own good. Is there any way to test that? Is there any way to separate network latency from the box being slow, or perhaps just plain busy restoring a backup to fix up the most recent crash of some piece of hardware or who-knows-what software?
Something that periodically runs a lightweight benchmark and logs the results comes to mind. Is there something that is commonly used for such measurments? I would rather not add too badly to machine's workload, but it gets sloooooowwwwwww at times :( Getting some data to objectify that statement could be useful.
Bill
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
-- = = = ======================================================================== John M. McIntosh <johnmci@smalltalkconsulting.com> Twitter: squeaker68882 Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com = = = ========================================================================
Thanks! I was thinking in terms of "anything but Smalltalk" to avoid the obligatory FUD and Mud slinging. However, it does sound like a useful idea, and if the code is already written, it can't hurt to see what it shows. If it flags something, then I can figure out whether I can get the IT types to believe it. Bill -----Original Message----- From: pharo-project-bounces@lists.gforge.inria.fr [mailto:pharo-project-bounces@lists.gforge.inria.fr] On Behalf Of John M McIntosh Sent: Monday, September 28, 2009 6:50 PM To: Pharo-project@lists.gforge.inria.fr Subject: Re: [Pharo-project] OT: vm performance There was code supplied by both Andreas and Bert and corrected by Juan that addressed the issue of profiling a server image Subject: Re: [squeak-dev] Re: How to profile a server image? Date: March 4, 2009 5:07:52 AM PST Perhaps you have a seaside component that kicks off a MessageTally then it shows you the results to determine if things look correct. You'll need to follow the emails and see if any of the changes actually made it into Pharo, it would have been helpful if they were. You could use OSProcess to pull the results of the command 'uptime' so you can understand how busy it is. On 2009-09-28, at 4:37 PM, Schwab,Wilhelm K wrote:
Hello all,
Here I am referring to a vm in the OS sense. Suppose you are forced to run code on a virtual machine running on a blade computer hidden behind locked doors. Access to the machine is limited to a remote desktop connection. You suspect that the machine is far too busy for its own good. Is there any way to test that? Is there any way to separate network latency from the box being slow, or perhaps just plain busy restoring a backup to fix up the most recent crash of some piece of hardware or who-knows-what software?
Something that periodically runs a lightweight benchmark and logs the results comes to mind. Is there something that is commonly used for such measurments? I would rather not add too badly to machine's workload, but it gets sloooooowwwwwww at times :( Getting some data to objectify that statement could be useful.
Bill
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
-- = = = ======================================================================== John M. McIntosh <johnmci@smalltalkconsulting.com> Twitter: squeaker68882 Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com = = = ======================================================================== _______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Well you can think different, some links below point to a status screen I did for a VisualWorks server application Red varies from 0-100 % and the sum of the other colors Green, Blue, Light Blue, and Purple are measured as state changes as work progressed for 4 different processes. http://smalltalkconsulting.com/clientSupport/L/opsguide/Image18.gif http://smalltalkconsulting.com/clientSupport/L/opsguide/Image17.gif http://smalltalkconsulting.com/clientSupport/L/opsguide/Image16.gif I think... goodness that was 12 years back, I did something like... footaskcounter := footaskcounter + 10. foo msg1 footaskcounter := footaskcounter + 10. foo msg2 footaskcounter := footaskcounter + 10. foo msg3 footaskcounter := footaskcounter - 30. As you can see from the gif files it was easy to look at the visual and understand what the application was *doing*. The support team could look at Image18.gif and fully understand what was happening. Operations only knew that Image18.gif was *normal* and that Image16.gif was the loss of the mainframe network data feed connection,. and for Image17.gif the image was doing nothing useful, idle, but not insane... On 2009-09-28, at 6:24 PM, Schwab,Wilhelm K wrote:
Thanks! I was thinking in terms of "anything but Smalltalk" to avoid the obligatory FUD and Mud slinging. However, it does sound like a useful idea, and if the code is already written, it can't hurt to see what it shows. If it flags something, then I can figure out whether I can get the IT types to believe it.
Bill
-- = = = ======================================================================== John M. McIntosh <johnmci@smalltalkconsulting.com> Twitter: squeaker68882 Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com = = = ========================================================================
I was going to let Wilhelm cross check the squeak list messages about message tally from earlier this year then hopefully get back to us On 2009-09-28, at 10:56 PM, Stéphane Ducasse wrote:
On Sep 29, 2009, at 1:50 AM, John M McIntosh wrote:
You'll need to follow the emails and see if any of the changes actually made it into Pharo, it would have been helpful if they were.
which changes could be added to pharo john?
-- = = = ======================================================================== John M. McIntosh <johnmci@smalltalkconsulting.com> Twitter: squeaker68882 Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com = = = ========================================================================
ok I integrated the latest changes of Juan (he fixed some not consistent behavior and cleaned the code). Stef On Sep 29, 2009, at 8:11 AM, John M McIntosh wrote:
I was going to let Wilhelm cross check the squeak list messages about message tally from earlier this year then hopefully get back to us
On 2009-09-28, at 10:56 PM, Stéphane Ducasse wrote:
On Sep 29, 2009, at 1:50 AM, John M McIntosh wrote:
You'll need to follow the emails and see if any of the changes actually made it into Pharo, it would have been helpful if they were.
which changes could be added to pharo john?
-- = = = = = ====================================================================== John M. McIntosh <johnmci@smalltalkconsulting.com> Twitter: squeaker68882 Corporate Smalltalk Consulting Ltd. http:// www.smalltalkconsulting.com = = = = = ======================================================================
participants (3)
-
John M McIntosh -
Schwab,Wilhelm K -
Stéphane Ducasse