In many instances you can accomplish the same task using either a stored procedure or a function. Both functions and stored procedures can be custom defined and part of any application. Functions, on the other hand, are designed to send their output to a query or T-SQL statement. For example, User Defined Functions (UDFs) can run an executable file from SQL SELECT or an action query, while Stored Procedures (SPROC) use EXECUTE or EXEC to run. Both are instantiated using CREATE FUNCTION. UDFs are instantiated using CREATE FUNCTION and SPROC instantiated by using CREATE PROCEDURE.
To decide between using one of the two, keep in mind the fundamental difference between them: stored procedures are designed to return its output to the application. A UDF returns table variables, while a SPROC can't return a table variable although it can create a table. Another significant difference between them is that UDFs can't change the server environment or your operating system environment, while a SPROC can. Operationally, when T-SQL encounters an error the function stops, while T-SQL will ignore an error in a SPROC and proceed to the next statement in your code (provided you've included error handling support). You'll also find that although a SPROC can be used in an XML FOR clause, a UDF cannot be.
If you have an operation such as a query with a FROM clause that requires a rowset be drawn from a table or set of tables, then a function will be your appropriate choice. However, when you want to use that same rowset in your application the better choice would be a stored procedure.
There's quite a bit of debate about the performance benefits of UDFs vs. SPROCs. You might be tempted to believe that stored procedures add more overhead to your server than a UDF. Depending upon how your write your code and the type of data you're processing, this might not be the case. It's always a good idea to text your data in important or time-consuming operations by trying both types of methods on them.