High CPU, low I/O, how to achieve it?

One of my colleagues mentioned about improving server performance through programming with as least I/O as possible by reducing the use of SQL statement. I'm wondering, how can a program achieve that ? Is there any alternative without using SQL statement to access SQL Servers like 2005 and the upcoming 2008?

I know Cobol on AS/400 has no problem, it can access the data directly without going through SQL statements, and DB2 is able to manage the flat file like a database.

 Any idea?

maybe he's referring to minimising database reads by putting everything inside RAM so that the app doesn't have to hit the disk during processing?

sounds kinda vague to me and i have no idea too how to achieve it without creating your own db server software  

My opinion is rewriting SQL statements to leverage on indexes which helps to reduce I/O. That is why, almost every tuning activity in any database servers is ultimately to reduce disk I/O. When we tune an SQL statement to replace a full-table scan with an index range scan, the performance improvement is the direct result of a reduction in disk I/O.

Well, the only other thing I can think of similar to without using SQL statements is probably LINQ or SQL CLR (which is still the same). It probably would not be the same as Cobol or DB2, I guess, since I am not familiar with. But still, can they achieve the same performance for large database files through accessing data directly?

tingshen:

One of my colleagues mentioned about improving server performance through programming with as least I/O as possible by reducing the use of SQL statement.

What kind of examples did he quote? That seems like a rather wide-covering and vague description.

It is not possible to cache all database data in memory - there simply ain't enough of it in any hardware platform today. What you certainly can do is cache pockets of frequently used result sets that don't get changed often. The SqlCacheDependency class is one such mechanism to limit database calls.

http://msdn2.microsoft.com/en-us/library/ms178604.aspx
http://msdn2.microsoft.com/en-us/library/e3w8402y(VS.80).aspx

and using an index isn't always helpful. the values in the indexed column should be sufficiently unique with respect to the values in the other rows. for example if the column is a bit (0 or 1) it's useless to index. there should be sufficient variation among values in that column for an index to be useful.