A story of using Riak for its Dynamo-like building blocks:

I made use of riak-kv, riak-search, and riak-core to help distribute my application’s data, state, and processing, respectively. These technologies provided me with a framework I could use instead of trying to build the functionality myself. This allowed me to put the entire application behind a dumb splitter and call it a day. A horizontal scale out is a simple matter of standing up a new node, joining it to the cluster, and adding an entry to the splitter.

The usage of riak-search as a distributed file index is interesting, given there’s also Luwak — but that Riak library is meant for large files.

As the author mentions this is an internal application of AOL Advertising. The public facing AOL Advertising is using Membase and Hadoop.

Original title and link: AOL AdLead Powered by Riak Components (NoSQL databases © myNoSQL)

via: http://www.progski.net/blog/2011/aol_meet_riak.html

Combining Hadoop and Membase to solve these challenges:

1. How to analyze billions of user-related events, presented as a mix of structured and unstructured data, to infer demographic, psychographic and behavioral characteristics that are encapsulated into hundreds of millions of “cookie profiles”
2. How to make hundreds of millions of cookie profiles available to their ad targeting platform with sub-millisecond, random read latency
3. How to keep the user profiles fresh and current

In a much simplified form:

• crunch (nb: read it as pre-process and prepare) tons of data with Hadoop
• feed the results in a low latency, high throughput key-value store for serving them online

Original title and link: Hadoop and Membase Case Study: AOL Advertising Architecture (NoSQL databases © myNoSQL)

via: http://www.cloudera.com/blog/2011/02/an-emerging-data-management-architectural-pattern-behind-interactive-web-application/