It is expensive to create an SQL connection, a socket connection must be established, a handshake must occurred, the connection credential must be check against the list of known credentials. To optimize this a technique called SQL Connection Pooling is implemented.
Every time a connection is required, we request it from the connection pool, if there’s exist the connection by the specified connection string then it’s going to return that connection rather than creating a new one. Every time we are going to close a connection we do not close it but instead we returned it to the connection pool. This way we can save the connection for later use.
reference: http://msdn.microsoft.com/en-us/library/8xx3tyca(v=vs.80).aspx
Rey Rahadian 