Lars Aronsson wrote:
The developers seem totally clueless about performance issues. There is a constant focus on new functions, and none on performance and response times. They cannot even say what the average response time has been in the last week or 24 hours. (Hey, prove me wrong: Tell me the average.)
I don't think this is accurate. The blame for the performance problem rests squarely on my shoulders, not on the shoulders of the developers.
On the current machine, Jason and I have login accounts on the machine, but none of the developers do. This makes it virtually impossible for them to do serious study of performance issues. They have a lot of clues, and a lot of suggestions, but no means to implement them -- I have been a bottleneck to the process.
This is all about to change, this upcoming Saturday. The new machine is wikipedia-only, and developers already have access to the machine. Root access is available for a small handful, even.
This means that the developers can sit on the "live machine" and study and tweak for performance, without having to guess what is wrong, and with instant feedback possible, without waiting for me to get around to fixing things.
It would be unfair for me to let the developers take the blame for a problem caused by a single-point-of-failure that is me! :-)
--Jimbo