High performance distributed data storage systems

George Bagropoulos — Sat, 19 Sep 2009 14:33:00 +0000

In the online edition of ACM Communications Sept. 2009 Issue I read a very interesting article from Michael Stonebraker’s blog about the end of RDMSs. I did a little research and I found that the distributed key-value stores are gaining ground especially in Web based applications. Many sites such as Amazon, LinkedIn, Yahoo, Facebook use distributed key – value stored to cope with petabyte data.

From the braindump blog I saw that there is a fast growing noSQL community that talks about how distributed non relational databases work.

A very interesting paper is the presentation of Amazon’s Dynamo highly available key-value storage system. From Amazon’s experience we learn that the reliability and scalability of a system is dependent on how its application state is managed.

For some quick links:

Hypertable is designed to manage the storage and processing of information on a large cluster of commodity servers, providing resilience to machine and component failures.

Cassandra is a highly scalable, eventually consistent, distributed, structured key-value store.

Voldemort is a distributed key-value storage system. In the “Build and Break” blog there is a nice post about the Voldemort project.

CouchDB is a document database server, accessible via a RESTful JSON API.

Mongo (from “humongous”) is a high-performance, open source, schema-free document-oriented database.

noSQL – Tales from the bits

High performance distributed data storage systems