Ask HN: Are you using Riak? What for? | Hacker News Reader

Ask HN: Are you using Riak? What for?

65 pointssimonw16 comments

At dotCloud we use PostgreSQL, Riak and redis.

- PostgreSQL for the user database since it maps nicely to django and sqlalchemy models

- Redis for caches and non persistent data

- Riak for everything else

We love the fact that it is easy to scale, that bitcask is safe (it writes to disk:-p) and easy to backup (rsync or cp is enough).

I'd be interested in hearing some thoughts on backing up. Would rsync to S3 happening regularly be an adequate solution? Figure if it is on a cron and it fires every other minute, the deltas would be pretty small, and not much would be lost if a datacenter or cluster goes down.

At Yammer we use Riak to power our notifications backend. More details available here…

http://blog.basho.com/2011/03/28/Riak-and-Scala-at-Yammer/

We use Riak at Dropcam to store H.264 video streaming from cameras running 24/7 which needs to be available for playback on the website and mobile apps. We needed a block store that could scale to handle petabytes, tolerate disk and node failures, and provide always-on write- and read-ability.

Riak fits the bill very well, though we did need to make a few mods to workaround certain problems. But it's Erlang/OTP, fairly small code-wise, and easy to modify.

Do you use luwak to store the videos or are they small enough?

It's a continuous stream. We break it into chunks similar to what Luwak does, though.

We use Riak as our primary datastore at Showyou, with Redis, Solr, and Memcached helping out. The big advantage of Riak is high availability and easy management, but for sorting/listing keys, range queries, and low latency we need other datastores which maintain post-commit-hook updated indices.

Out of curiosity, did you try riak_search as a solr replacement?

Riak provides post-commit hooks:

http://wiki.basho.com/Pre--and-Post-Commit-Hooks.html

But I assume you mean these post-commit hooks update other indices stored outside of Riak.

Yes, sorry, the wording was somewhat ambiguous. We use post-commit-hooks in riak to update external datastores, as well as callbacks within our ORM for many indexes stored in Riak itself.

There was recently a poll:what database your company is using? Which within the commentd riak was mentioned mant times

http://news.ycombinator.com/item?id=2684620

Don't know why that got downvoted, since it's a link to relevant prior discussion. But here's a more specific link (to the Riak thread in that discussion): http://news.ycombinator.com/item?id=2685053

For now, nothing. My systems use redis as a datastore at first, then move to riak when they outgrow redis. So far my tiny systems haven't outgrown redis.

When I have time, I'll add distributed counter support to Riak and take over the world. (Let's count everything. In real time with massive concurrency, full historical resolution, and no single points of failure.)

2 Ruby client projects. Really happy with it. always use it with Redis so we can get atomicity for things like counters, etc...

Using it to power the activity stream of a small site (+ Postgres & Redis). <3

Reply on news.ycombinator.com