A cube can be stored on a single analysis server and then defined as a linked cube on other Analysis servers. End users connected to any of these analysis servers can then access the cube. This arrangement avoids the more costly alternative of storing and maintaining copies of a cube on multiple analysis servers. linked cubes can be connected using TCP/IP or HTTP. To end users a linked cube looks like a regular cube.
A cube can be partioned in 3 ways.Replicate,Transparent and Linked.
In the linked cube the data cells can be linked in to another analytical database.If an end-user clicks on a data cell ,you are actually linking through another analytic database.
Partitioning a cube maily used for optimization.(ex) U may have data for 5gb to create a report u can specify a size for a cube as 2gb so if the cube exceeds 2gb it automatically creates the second cube to store the data.
0 
1