On Sun, Oct 25, 2009 at 3:40 PM, Michael Miller <[email protected]> wrote: > Hi All, > > I should introduce myself to the list: I am one of the cofounders of > Cloudant. Behind Adam Kocoloski's lead we are actively contributing to > CouchDB and are currently deploying it at scale as a hosted service for our > alpha-customers. We are co-sponsoring this week's no:sql(east) conference > in Atlanta (*) and I'll be giving my first talk on CouchDB. The talk is > aimed squarely at telling the story of couch in the wild, and I could use > your help. After a brief description of couch and what makes it unique, I > want to focus on what it's like to use couch in production. In particular: > > - why did you choose couch? We needed to store tons of facts/profiles about millions of users. These profiles are 'loosely' schema-ed, and we needed an efficient way to query them. As such, storing them in a relational db made little sense. I had played with document databases(Couchdb) a little and they seemed like a better solution to our problem but we didn't want to rely on amazon for this feature. Couchdb seemed interesting.
> - if you did not choose couch, what was the leading reason? There are still a few places where we'll need uniqueness constraints on more than one field. > - what problem(s) does couch solve for your application? Fast, query-able datastore that can scale up to hundreds of millions of documents. Replication makes writing enabling failover seamless. It's gotten so popular @ meebo that a lot of our internal dashboards are written as couch-apps (our JS developers love that map/reduce is implemented in JavaScript). > - was couch a success for your needs? Absolutely. We're moving to a model where all of our data that would be stored on a sharded mysql instances will be stored in couch instead. > - what are your top feature requests for future development? Speed, stability. Rolling the features of the lounge into couch (especially the smartproxy). Filtered changes (I think this is scheduled for .11 right?) > - what do you love about couch? Scheme-less design means that we don't have to alter tables when things inevitably change. Views allow us to query the data the exact way we want it. Replication gives us the ability to > - what do you hate about couch? Now I need to learn erlang. > - AOB > > I realize a lot of this information is available from the most recent survey > (**) but there has been a lot of action since that was completed. Any > feedback that you can provide would be greatly appreciated. > > Thanks, Mike > > -- > Mike Miller > [email protected] > > > (*) https://nosqleast.com/2009/ > (**) http://survey.io/results/842d257419c0aab16450cb00cf930a99df7c4257 >
