Managed Code in SQL advantage confusion (C#)

[manyuaditya]

Hi,

I'm developing this scientific application that does the following-

1. read a data table ( 100's of millions of rows )
2. Create a tree out of this data ( a kd-tree which is sort of like a binary search tree but over multiple dimensions )
3. store this tree in a database table

Part2

1. Read a previously stored tree into memory
2. Read the data table that was used to create the tree
3. now performs some complicated tree traversal etc on this data set and tree
4. store results of traversal into a database table.

Now I have written the application entirely in C#. the initial plan was to create an assembly out of this code and put it inside SQL server so that it can run in SQL CLR and then users can call this function through query analyzer and it will do everything from loading data to storing table.

Now, suddenly I'm a bit confused. The only code that I have to change in my application to put it into SQL server as managed code is where i get my connection I use the connection string to get the current context connection...

The rest all seems exactly like a typical program that runs outside of sql server. I get a connection .. read the data into an array in memory, do computation, then write to a database table. So I'm failing to see here the advantage of importing the code into sql server. In fact I see the disadvantage that if I have a table with 100's of millions of row's I shouldn't be asking for sql server for so much memory anyway.

Am I missing something. Have I approached the application's development in a wrong manner? Will there be an advantage ( performance wise ) if I run this inside SQL server.