Site Scalability with Cache::Memcached

bageler has asked for the wisdom of the Perl Monks concerning the following question:

I have used Cache::Memcached experimentally and love it, but I was wondering if any of our esteemed monks have extensive experience using memcached. I'd like to know about your experiences with it have been like, any pointers you might have, and what you would think if there was a Tie::Hash::Memcached module (which I am seriously contemplating writing tomorrow). Thank you sirs and ladies!

Comment on Site Scalability with Cache::Memcached

Replies are listed 'Best First'.
Re: Site Scalability with Cache::Memcached by perrin (Chancellor) on Dec 08, 2005 at 02:18 UTC
Since LiveJournal runs it, I think you can rely on it being scalable. A tie module is probably not a good idea: you will add overhead to your cache calls which defeats the purpose of having the cache. (Method calls are faster than tied hashes.)	[reply]
Re^2: Site Scalability with Cache::Memcached (proportion) by tye (Sage) on Dec 08, 2005 at 09:45 UTC
No, the point of memcached is that fetching simply keyed data from a special-purpose daemon can be tons faster than fetching from a database. A tied interface is also tons faster than fetching things over a network. So the overhead of the former (compared to method calls which are slower than plain function calls which are slower than C code) is not at all in conflict with the purpose of memcached. - tye	[reply]
Re^3: Site Scalability with Cache::Memcached (proportion) by perrin (Chancellor) on Dec 08, 2005 at 13:46 UTC
It's certainly a step in the wrong direction, if fast lookups are what you're after. Many people imagine that TIE will make things faster, so pointing out that it makes things slower seems useful to me.	[reply]
Re^2: Site Scalability with Cache::Memcached by bageler (Hermit) on Dec 08, 2005 at 20:43 UTC
The Tied interface point was a separate thought from scalability. I'm mainly looking for personal experiences with the package, I can read the internet as well as the next person to see that livejournal et al. use it.	[reply]


Perl: the Markov chain saw
	PerlMonks