Thoughts on SCADS: Scale-Independent Storage for Social Computing Applications

Authors: Michael Armburst, Armando Fox, David Patterson, Nick Lanham, Beth Trushkowsky, Jesse Trutna, Haruki Oh

BibTeX:
@proceedings { citation185,
title = {{SCADS}: Scale-independent storage for social computing applications},
year = {2009},
month = {01/2009},
booktitle={Conference on Innovative Data Systems Research {(CIDR)}}
URL = {http://www-db.cs.wisc.edu/cidr/cidr2009/Paper_86.pdf},
author = {Michael Armbrust and Armando Fox and David Patterson and Nick Lanham and Haruki Oh and Beth Trushkowsky and Jesse Trutna}
}

Summary:
The authors are working on a framework that allows developers to scale up as well as scale down easily. There are three main parts to this: A query language that does not compromise performance with scale, a declarative policy that allows developers to specify performance levels that the framework should achieve, and machine learning algorithms to rapidly scale up and down. One of the main points is that consistency can be traded for performance.

The Good:
The system looks cool. It would be nice to have the performance of memcached with a flexible query language. It's nice to throw all the load on some machine learning algorithm to scale up and down.

The Bad:
System is still preliminary.

The Ugly:
This seems a huge amount of work. I wonder if it will ever work.