Re: [codesprints] Performance oriented codesprint before IETF 112

[Robert Sparks <[e-mail]>]

Given the doodle results, my plan is to have the primary sprint 27 Oct 
1600-2000 UTC. People can join or leave as convenient. Coordinates will 
be sent soon.

I will also have a conference set up for 28 Oct at those same times for 
followup and for anyone who can't participate on the 27th.

I plan to make resources for analyzing performance available well before 
the sprint. Future communication on that topic will occur on the 
codesprints@ list. (If you are interested, please subscribe).

RjS

On 9/29/21 5:23 PM, Robert Sparks wrote:
> We would like to hold a codesprint before IETF 112 focused on 
> datatracker performance.
>
> We have been improving the instrumentation of the datatracker, and 
> plan to provide quite a bit of information that will help us focus on 
> places that will help us most. These include statistic from the 
> database engine itself, and instrumentation that measures which views 
> are getting the most use, suffering the worst cumulative performance, 
> and how those views use the database.
>
> I expect much of the work will be verifying that the database has the 
> right indices, adjusting views to make better use of the ORM (such as 
> prefetching data (and not to prefetch too _much_ data), and adjusting 
> models where the current models are proving to be problematic.
>
> I have a doodle for two different possible times on either Oct 27 or 
> Oct 28.
>
> If you would like to participate, please add yourself to the doodle. 
> If the proposed times can't be made to work for you and you would 
> really like to engage in this kind of sprint, please just add a 
> comment at the doodle page.
>
> RjS
>
>
>
>