Question

I’m writing a command-line tool (to run on Windows 7 boxes) that inserts a single record into a table and then exits. I’m using C# (.NET 4.5) and SQL Server 2008.

The record of data contains a large (up to 6000 bytes) string, a time-stamp and an auto-incrementing integer ID primary key. There are no foreign keys attached to the table.

Also, since this is the most common usage scenario for this database I’d like to optimise for it. Are there any tricks with how I set up the database that will help keep individual insert times low?

[EDIT]
The code I’m currently using…

SqlConnection con = new SqlConnection(ConnectionString);
con.Open();
SqlTransaction trans = con.BeginTransaction();

SqlCommand cmd = new SqlCommand("INSERT INTO [msglog](message,project,username,computername,updated_at) VALUES (@m,@p,@u,@c,@i,@t)", con, trans);
cmd.Parameters.AddWithValue("@u", Username);
cmd.Parameters.AddWithValue("@c", Computername);
cmd.Parameters.AddWithValue("@m", msg);
cmd.Parameters.AddWithValue("@p", ProjectName);
cmd.Parameters.AddWithValue("@t", DateTime.Now);
cmd.ExecuteNonQuery();
trans.Commit();

Answer 1

In your code example, I suspect that most of the time is spent establishing a connection to SQL Server. I would recommend:

• Attempt to reuse connections / a single instance of your application. Your question suggests that the application opens, inserts a record, and then exits. This requires constantly re-establishing a new connection for a single statement, not to mention overhead in launching the application itself.
• Remove your transaction (it’s not needed for a single statement).
• Don’t bother trying to build a single string for your statement. You should use parameters, and it will likely be faster to use parameters. SQL Server is more likely to cache the execution plan for identical queries (with changing parameter values).
• Make your table inserts fast (RedFilter has some great suggestions here). I’d add that you would ideally cluster your table by an incrementing key, so that inserts are always at one end of the table.

There may be other micro-optimizations, but really addressing my first concern above would give you the most significant speedup.