9 comments

[ 3.2 ms ] story [ 32.7 ms ] thread
Somehow I was expecting an article on statd, part of NFS.
Funny, me too, given how I was just researching server side locking and sqlite yesterday.
Also check out Statsite, which is a C implementation of StatsD: https://github.com/armon/statsite
Yup! We actually linked to that from the article as an example C implementation. It's really solid and also has a Librato backend ;-). We're already using it on a several high-throughput services internally and it will probably warrant it's own post in the future.
Thanks for this! Looks really promising. I'm a huge fan of statsd, but it seems like something so much better suited to a single binary than a whole node setup.
So ... rrdtool?
We used that for a long time, but it stopped scaling at some point. We ended up building IMVU istatd, the highest performance real time counter database we know of. Github.com:imvu-open/istatd Could be a back end for statd, or an alternative entirely; your choice.
What's the pricing like for Librato?
It's AWS-style metered and billed monthly based on the number of time-series or "data streams" you're sending us.