.NET Applications: Writing Good Code - The Data Access - page 215

• The DataReader object is optimized for retrieving large amounts of data at a

fast speed. Data is read-only and can only be fetched in a forward-only order.
Memory usage is minimal.

• The DataSet object is a cache of data that represents the complete set of data

results, including related tables, constraints, and relationships among the
tables. It is, in effect, a locally cached database. You can modify the data in
the DataSet and fetch data in any order. Because the DataSet is discon-
nected from the database, any changes you make to the data in the DataSet
must be explicitly synchronized with the data in the database. You also can
create a DataSet from an XML stream or document, or you can serialize a
DataSet to XML. Memory usage is high.

Performance Tip

If you need to retrieve large amounts of read-only data, a DataReader
object always provides the best performance. Only use a DataSet object
if you need to insert, update, or delete data, fetch data in any order, or
work with XML. Although the flexibility of a DataSet can benefit your
application, it comes with the high cost in memory consumption.

Using GetXXX Methods to Fetch Data from a DataReader
The .NET API provides the following methods of fetching data from a

DataReader :

• Generic data type methods, such as GetValue() and GetValues()

• Specific data type methods, such as GetDateTime() , GetDecimal() , and

GetInt32()

When using a generic method such as GetValue() to fetch data from the
DataReader , extra processing is required to convert the value data type to a ref-
erence data type, essentially wrapping the value data type with an object. This
process is called boxing . When boxing occurs, memory is allocated from the
managed heap on the database client to create an object for the reference data
type, which can force a garbage collection to occur. See “.NET CLR,” page 82, for
more information about the impact garbage collection has on performance.