CareerCup

1) Partition of database
a) Different tables to different servers
b) Divide large table into multiple partitions to different servers
2) Partition of queries accordingly
3) Develop query cache system to avoid frequent query of database with identical conditions.

Lets assume that single data server can handle one million request. we need 100 data server . same time we need to distribute this data on 100 server. create partition table so that request can routed to correct server.

Well I think he is looking for Map Reducer logic.
It has two phase.

map Phase:
Say if you have n computer then use (n-1) to serve pages each capable of storing say 1 million results.

reducer Phase:
It maintains a list of all the request from all the n-1 server but there is not need to join the result.
When ever a request is place just forward it to the appropriate server. :)