Basho (creators of Riak, growing quickly) opens SF office

blog.basho.com

30 points·roder··6 comments

Mozilla is using it for our Test Pilot back-end. Note that vector clocks are part of the original Amazon Dynamo paper.

It is pretty cool. One of the big wins was being able to submit MapReduce jobs written in JavaScript through a REST API. The other was the fact that there is no server middleware in this application. just the REST API of Riak.

We should have some benchmarking and field testing information up on our blog soon.

Biggest limitation we've hit so far is that it is very inefficient to try to run a MapReduce that scans over every key in a bucket when you have lots of buckets with billions of items. In the current versions of Riak, you really need to explicitly pass the M/R a list of bucket/keys that you want to work on.

http://blog.mozilla.com/data/2010/05/18/riak-and-cassandra-a...

The key scanning is a major limitation but is offset if you keep separate key indices elsewhere or in specific known keys in riak. If you can deal with that and no ordering then riak is a great choice for a distributed platform.

Has anyone used Riak? The vector clock seems interesting, and I'm wondering how it works out in practice.

the vector clock is practically transparent unless you actually want to take advantage of it.

I love their "drop by anytime" invitation.

I'll see you Wednesday afternoon, Basho West.

We look forward to it! We have Riak t-shirts too.

We've also just started the San Francisco Riak Meetup: http://www.meetup.com/San-Francisco-Riak-Meetup/ .

Reply on news.ycombinator.com