getting near real time request counts

[Matthew M. Boedicker]

> How close to realtime do you need?  What sorts of grouping are you wanting
> (e.g. the number of requests in the last minute)?  How long do you need to
> store this data?  Are you ok with losing precision as time passes, as happens
> in rrdtool and friends?  Are you concerned with all urls, or only certain ones,
> or only the top x%?  How important is it to you to not lose any data (is it ok
> to send stats around via UDP)?

I had the same concern about slowing down responses, so it would have
to be very lightweight. Maybe UDP as you said or putting something
into a queue.

I want to track total hit counts of image urls to track their
popularity. A single value per url is good enough and it can be pretty
far from real time.

Almost all of them are cached in varnish so varnish is the only thing
that knows that they have been requested.