JDBC Performance is no where near .NET performance (JDBC 1..

I am finding that the adaptive buffering returns rows at the same rate as
the server side cursor. It may improve client memory footprint on large
resultsets but I cannot seem to get performance up to the level of a similar
app written in .NET using SqlDataReader classes. The JDBC performance is
really bad. This only seems to be an issue in the large resultset scenario
which is what we deal with in our product.

My test app creates 6 threads, each with their own connection to the same
SQL Server instance, query different tables with the same replicated data.
The resultset will return 83,000 rows of 17 columns of data (a good mix of
data types...no blobs). The app will loop through every row and column but
do nothing with the data, except reading it.

JDBC averages 79.0 seconds
..NET averages 0.90 seconds

Any ideas?

Here is a snippet of the java thread test code:
....
Connection c = DriverManager.getConnection("jdbc:sqlserver://" + host +
":1433;databaseName=" + database + ";user=" + user +
";password=" + password +
";responseBuffering=adaptive;");

// Statement stmt = c.createStatement();
Statement stmt =
c.createStatement(SQLServerResultSet.TYPE_SS_SERVER_CURSOR_FORWARD_ONLY,
SQLServerResultSet.CONCUR_READ_ONLY);

((SQLServerStatement) stmt).setResponseBuffering("adaptive");
ResultSet rs = stmt.executeQuery(sql);
rs.setFetchSize(10000);

ResultSetMetaData rsMetaData = rs.getMetaData();
int numberOfColumns = rsMetaData.getColumnCount();

String tempVar;
int rowCounter = 0;

// loop through rows in resultset
while (rs.next()) {
// loop all columns and read value
for (int i=1; i<numberOfColumns; i++) {
tempVar = rs.getString(i);
}

rowCounter += 1;
}
....
Is the JDBC code that much different at the TDS protocol than .NET? Any
suggestions to improve this would be great!

Thanks

Hi,

If you use a server cursor then adaptive buffering is defeated except with
forward only updatable result sets. But why are you using a server cursor
here? Your code looks like it just needs a standard default forward only
read only result set. That is, it should be sufficient just to create the
statement with:

Statement stmt = c.createStatement();

In that case, you also don't need to set adaptive buffering at the
Statement level, since you already set it in the connection properties.

Hope this helps (let me know if it doesn't),
--David Olix [SQL Server]