Arm-waving approach to the issues and approaches
================================================

Issues
 - how to quickly and reliably handle the mass of communication 
   taking place, from all points of the globe, with wildly varying 
   and often unpreditable loads, 

 - how to store/retrieve/update the mass of information needed
   for the particular application

   e.g. amazon - huge volume of salable goods,
                 huge volume of requests,
                 financial transactions need to be reliable
        twitter - huge volume of time-sensitive communication,
                 ensuring tweets reach all the intended followers
                 without overwhelming data centre or networks
        facebook - huge volume of information stored, heavy
                 posting/querying loads
        youtube - huge volume of information to be stored,
                 searched, and retrieved

Approaches

 - use traditional network/rdbms approaches, just go for
   more and better hardware (rapidly gets $$$ EXPENSIVE $$$)

 - replication: have data copied in many places, users
   get directed to one of them either randomly or by geography
   (problem is keeping the copies synchronized: see sharding)

 - go to non-traditional dbms, typically key-value pairs,
   designed for fast access (problem is limited query types
   or high-end processing before/after db query)

 - server everything out of memory, keep your dbms primarly for
   long term storage/retrieval (e.g. twitter)