Recently I came to know about graph databases.I read that these databases have limited analytics. I read that here http://www.readwriteweb.com/enterprise/2009/02/is-the-relational-database-doomedp2.php "Things like tracking usage patterns and providing recommendations based on user histories may be difficult at best, and impossible at worst, with this type of database platform."
1 I am not able to understand why is this analytics is limited here?
2 How these database graph can be used for archiving for example facebook which saves all the posts by millions of users. How this can be done in graph database?
if you apply a strict Property Graph Model, you will find that you have great "data local" operations, like exploring the surrounding data of a node say, 5 hops deep along the relationships. However, global operations like "give me all nodes that have a name attribute of value 'Tom*'" require in a graph model a full scan of the data. This is in theory a limit. In practice (like in http://neo4j.org) the graph engine is combined with global indecies like Lucene, BerkelyDB or Cassandra, that can take care of this kind of data-global aspects that are often used in certain analytics scenarios.
So, there is no real limitation, just a different way and different patterns to deal with global and local operations of your data. See http://wiki.neo4j.org/content/Domain_Modeling_Gallery for some examples of modeling domains in graphs, or even GIS examples like https://github.com/neo4j/neo4j-spatial/raw/master/src/site/pics/one-street.png on the Open Street Map graph.
For archiving vast amounts of data like Facebook, I would only store say the last month or so of status updates for fast retrieval and recommendations in a graph. The rest I would archive in solutions like Cassandra and just have a reference and key metrics of the archive in the graph on how to retrieve this "archived subgraph" if needed.
