SQLBulkCopy performance problem

  • Thread starter Thread starter AdrianDams
  • Start date Start date
A

AdrianDams

I have a solution where I am receiving a stream of data from a device and I
want to insert it into a SQL Server table. I am using Ado.net 2.0 and the
SQLBulkCopy class.

My test data amounts to about 50 tags a second and I have configured my
system so that I can change parameters of my bulk write operation. I
appreciate that this is hardly a lot of data but my target system will be a
lot bigger. In short, though, I am disappointed with the results and hope
someone can point out how to improve them. I have been hoping that
SQLBulkCopy would be able to handle high throughput at low CPU loads, and
not only bulk bursts of data inserts.

The code I am using does the following:
1. Creates an Ado.net DataTable
2. Creates an instance of the SQLBulkCopy class
3. Sets the DestinationTableName property of the SQLBulkCopy class to the
destination SQL Server table (which has appropriate indexes)
4. Insert rows into the DataTable once a second
5. When the SQLServerInsertInterval fires, the WriteToServer method of the
SQLBulkCopy class is called to flush the buffer (DataTable) to the
destination SQL Server table

These are the results of my test. I'd really appreciate any help here.

The tests were performed on my laptop:

Pentium

1G RAM

2GHz

Win Xp Sp2

There were a total of 50 tags changing once a second on the input stream

In summary, these are the results:

data acquisition cycle Time
SQL Server Insert method
SQL Server insert interval (SQLBulkCopy.WriteToServer)
Total Average CPU load
Cpu load used by SQL Server

1 second
Normal Insert
1 second
~70%
27%

1 second
Bulk Copy
1 second
~55%
~15%

1 second
Bulk Copy
Never (disabled)
~55%
~16%

1 second
Bulk Copy
60 seconds
~55%
~17%


In short, SQL Bulk Copy seems to give an improvement over normal inserts on
the total and SQL Server CPU load, but tweaking the parameters of the SQL
Bulk Copy operation doesn't seem to make much difference. In particular, the
fact that it seems to chew up the same amount of CPU even if there are no
inserts to the SQL Server is very hard to interpret.
 
SQLBulkCopy (like the BCP utility) has some overhead, but it's designed to
handle a tremendous volumes--I would not expect it to be that much faster
with just a few rows.

I just wrote an example program that moves 450,000 rows to the server in
under 30 seconds. I would run your tests with larger rowsets to get a feel
for the performance. In the past we also turned off the indexes on the
target table to further improve performance. Once the data was imported, we
executed SQL to re-establish the indexes.

--
____________________________________
William (Bill) Vaughn
Author, Mentor, Consultant
Microsoft MVP
www.betav.com/blog/billva
www.betav.com
Please reply only to the newsgroup so that others can benefit.
This posting is provided "AS IS" with no warranties, and confers no rights.
 
I'm not sure I understand your test. When you say that you don't send any
rows to SqlBulkCopy, where is it consuming CPU then? Also, I assume you
empty the DataTable (or create a new one) between calls to WriteToServer,
right?

--
Pablo Castro
Program Manager - ADO.NET Team
Microsoft Corp.

This posting is provided "AS IS" with no warranties, and confers no rights.
 
Back
Top