Long term problems issues

• We need some kind of caching system, for instance for liking system and for many more. Writing neo4j everytime is plain stupid. Neo4Django would be a natural candidate, but apparently it is on Gremlin and really, really slow.. We could modify it for our needs - someone has to think about it how to use it

• Relying on Neo4J might be a mistake. We have to pay for it.. There are other options. Well facebook doesn't use at all graph database like Neo4J, and everything in facebook is a graph (one tech talk I remember).. http://royal.pingdom.com/2010/06/18/the-software-behind-facebook/

• Django can be efficient : http://blog.disqus.com/post/62187806135/scaling-django-to-8-billion-page-views

• It is a good idea to see at the very beginning architectural problems of big web services : http://highscalability.com/blog/2013/7/8/the-architecture-twitter-uses-to-deal-with-150m-active-users.html

• I did something like this during my internship (not it, internal architecture very simillar to this one) : http://blog.samibadawi.com/2013/04/akka-vs-finagle-vs-storm.html .

• Comparision of graph topological paraller programming architectures : http://blog.samibadawi.com/2013/04/akka-vs-finagle-vs-storm.html