Skip to content

Instantly share code, notes, and snippets.

@PharkMillups
Created March 11, 2011 23:30
Show Gist options
  • Save PharkMillups/866802 to your computer and use it in GitHub Desktop.
Save PharkMillups/866802 to your computer and use it in GitHub Desktop.
13:16 <atoi> Any thoughts on Riak for stats for an ecommerce website?
Particularly when to keep the load on PostgreSQL low but still do
reporting? I'm assuming this is a decent use case, but...
13:23 <seancribbs> atoi: Riak is great for write-heavy, so if you
just needed to log something, it's a good fit.
13:25 <atoi> seancribbs, that's what I was thinking. My practical
test, I think, will be to implement some tasks that we'll need the
\stats system for and see how painful it is to get the alerts and
reports out of it we'll need.
13:26 <seancribbs> atoi: keep in mind the MapReduce is not (yet) tuned
for batch processing, so you may want to look into something else to handle statistics, etc.
13:26 <seancribbs> but just as a place to dump data reliably, it should
be great for you
13:27 <atoi> When you say batch processing, do you mean that, similarly
to Mongo, map/reduce is single process per box? or...
13:28 <seancribbs> atoi: no, i mean don't expect MapReduce on Riak to be Hadoop
13:28 <seancribbs> it's tuned for online queries over more limited datasets
13:29 <atoi> Any suggestions where I should start for batch
processing? Do you suggest just cooking something up homegrown?
13:29 <atoi> Because some of this will definitely be pulling a lot of
\stats for reports over large data sets.
13:53 <seancribbs> atoi: depends largely on your needs. (sorry for being distracted)
13:53 <atoi> no worries. ;)
13:53 <atoi> it's IRC
13:53 <seancribbs> one customer of ours uses Riak to feed into Hadoop
13:53 <seancribbs> but if you don't need to see the whole dataset you
could make a commit hook that does some processing
13:54 <seancribbs> we're looking into CQRS-style architectures for some applications
13:55 <atoi> Right. The main requirement is that we're going to be
stuffing a bunch of stuff in there but I'm not sure, ultimately, how we'll
want to use the metrics. Right of the bat, we'll be using it to set up alerts
and reporting on campaigns, similar to Google analytics campaigns. "Users should
go to A then B then C. If 100 users go to A but no one hits C, something could be
busted! Send an alert!"
13:56 <atoi> but ultimately we will almost certainly do a lot more than just that.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment