I am currently struggling with a problem bulk inserting data into sql server.
The application I am writing is multi-threaded and downloads about 2000 records every 2 seconds on x amount of threads (depending on bandwidth).
What I would like to do is find a 'friendly' way to insert this data into sql server without hammering the cpu.. I have tried the following with little success.
1, Using a single insert and thread.sleep(x * 20) to allow for massive data input, altough this made the application more stable and lowered cpu usage to very little the data takes about 60 times longer to download and process into the database.
2, Using a SqlDataAdapter and DataSet and updating the database via the .Update method of the data adapter.. Simply this was awful and took forever to process the data into the database.. (Took about 30 seconds to process 2000 records and Command Timeout was high).
3, Using OpenXML in SqlServer and parsing the data as an XML string (nText), although this method is fast its still very CPU intensive. I have to set the command timeouts very high to allow for this approach (because of the multi-threaded nature of the app).
Does anyone have any idea's on a cpu friendly approach to this problem ??
Thanks in advance..
Gary.You can actually use BCP
No comments:
Post a Comment