[Cosmo-dev] Cosmo performance

[Adam Christian]

I am currently measuring Cosmo UI performance using Windmill, there  
is a performance tab that uses a timing object I built in javascript  
to measure the length of time any action takes. In the near future we  
will be sending this back to the server and using python to store and  
analyze it, but for now I am working on a small ajax app to parse the  
current performance output and graph it so that we can compare  
performance times for each checkpoint we test with Windmill. As soon  
as I have a demo I will send out the url.

Adam


On Apr 4, 2007, at 5:52 PM, Ted Leung wrote:

> Hi folks
>
> At this week's preview countdown meeting we spent some time talking  
> about Cosmo performance.   With the switch to EIM/Morse Code, there  
> have been a lot of changes and we don't know if the baseline  
> performance numbers that Jared used to size the production server  
> machine still hold true.   Since one of the goals of the EIM/Morse  
> Code was to improve performance, we expect things to be better, but  
> we need to verify that.    Also, it appears, anecdotally, that  
> there are parts of the Cosmo UI that should be snappier than they  
> are.    Generally speaking, then, it seems that we probably need to  
> get more organized on the performance front.    To that end, I'd  
> like to make the following proposal for getting started:
>
> 1. Have Randy and Jared work together to put together some tests  
> that approximate the reasoning that used by Jared's old tests.    
> These tests were the ones that we used for a baseline initially.    
> Some of these tests need to be reworked because they are DAV based,  
> and we'll be using EIM based syncing now.
>
> 2. Start looking a ways to measure the performance of the Cosmo  
> UI.  In Chandler, we defined a set of use cases and a standard data  
> set, and measure the time it took for those operations to  
> complete.   Things are a bit more complicated with a multi-user web  
> app, but we ought to be able to at least get some raw numbers for  
> single user responsiveness as a start.   PPD is going to come up  
> with a list of use cases that we want to measure.
>
> Any other ideas for areas to look at or for methodology?
>
> Ted
> _______________________________________________
> cosmo-dev mailing list
> cosmo-dev at lists.osafoundation.org
> http://lists.osafoundation.org/mailman/listinfo/cosmo-dev