Why is OracleDataAdapter.Fill() Very Slow?

Question

I am using a pretty complex query to retrieve some data out of one of our billing databases.

I'm running in to an issue where the query seems to complete fairly quickly when executed with SQL Developer, but does not seem to ever finish when using the OracleDataAdapter.Fill() method.

I'm only trying to read about 1000 rows, and the query completes in SQL Developer in about 20 seconds.

What could be causing such drastic differences in performance? I have tons of other queries that run quickly using the same function.

---

Here is the code I'm using to execute the query:

using Oracle.DataAccess.Client;

...

public DataTable ExecuteExternalQuery(string connectionString, string providerName, string queryText)
{
    DbConnection connection = null;
    DbCommand selectCommand = null;
    DbDataAdapter adapter = null;

    switch (providerName)
    {
        case "System.Data.OracleClient":
        case "Oracle.DataAccess.Client":
            connection = new OracleConnection(connectionString);
            selectCommand = connection.CreateCommand();
            adapter = new OracleDataAdapter((OracleCommand)selectCommand);
            break;
        ...
    }

    DataTable table = null;
    try
    {
        connection.Open();

        selectCommand.CommandText = queryText;
        selectCommand.CommandTimeout = 300000;
        selectCommand.CommandType = CommandType.Text;

        table = new DataTable("result");
        table.Locale = CultureInfo.CurrentCulture;
        adapter.Fill(table);
    }
    finally
    {
        adapter.Dispose();

        if (connection.State != ConnectionState.Closed)
        {
            connection.Close();
        }
    }

    return table;
}

And here is the general outline of the SQL I'm using:

with
  trouble_calls as
  (
    select
      work_order_number,
      account_number,
      date_entered
    from
      work_orders
    where
      date_entered >= sysdate - (15 + 31)  -- Use the index to limit the number of rows scanned
     and
      wo_status not in ('Cancelled')
     and
      wo_type = 'Trouble Call'
  )
select
  account_number,
  work_order_number,
  date_entered
from
  trouble_calls wo
where
  wo.icoms_date >= sysdate - 15
 and
  (
    select
      count(*)
    from
      trouble_calls repeat
    where
      wo.account_number = repeat.account_number
     and
      wo.work_order_number <> repeat.work_order_number
     and
      wo.date_entered - repeat.date_entered between 0 and 30
  ) >= 1

Answer 1

There are known performance differences between using the Microsoft Data Provider for Oracle and the native Oracle Data Provider.

Have you tried both?

What are you trying to achieve with this query? Forget about technical stuff, just the objective of it all. Perhaps is there a tune possible for your query.

Have you tried with a profiler to see where it gets stuck?

Answer 2

I think the culture and the date being returned by your Oracle query are different and that is where the application is taking a lot of time to parse.

Answer 3

This code helped me, try it:

using (OracleConnection conn = new OracleConnection())
{
     OracleCommand comm = new OracleCommand();
     comm.Connection = conn;
     comm.FetchSize = comm.FetchSize * 16;
     comm.CommandText = "select * from some_table";

     try
     {
          conn.Open();
          OracleDataAdapter adap = new OracleDataAdapter(comm);
          System.Data.DataTable dt = new System.Data.DataTable();
          adap.Fill(dt);
     }
     finally
     {
          conn.Close();
     }
}

The trik is in line (try values from 8 to 64 to find the best for your case):

comm.FetchSize = comm.FetchSize * 16;

UPDATE:

Here is an improved code:

OracleConnection myConnection = new OracleConnection(myConnectionString);
OracleCommand myCommand = new OracleCommand(mySelectQuery, myConnection);
myConnection.Open();
using (OracleDataReader reader = myCommand.ExecuteReader(CommandBehavior.CloseConnection))
{
    // here goes the trick
    // lets get 1000 rows on each round trip
    reader.FetchSize = reader.RowSize * 1000;

    while (reader.Read())
    {
        // reads the records normally
    }
}// close and dispose stuff here

From here