Chung Leong wrote:
Load up the database with lots of test data and run through the site with
your browser. Perceived performance is what's important. After all, if an
improvement can only be detected by a test tool, what is the point?
Nothing if you have only one user. But if you got 10^X users, then wait
time for each user will increase, because they have to wait for other
viewers first.
For example, if it takes 1 second for the php to do it all, and we have
600 viewers. Assuming processor time will be shared equally for all,
then they all have to wait 10 minutes before they can see the page. If
time will be shared unequally, then first person might get the page in 1
second, but the last one would have to wait 10 minutes.
And now, if you can speed up the page into 0,5 seconds. Doesn't look
much, but viewers will see the page 5 minutes faster assuming we have
the same amount of viewers.