Protocol Encryption Performance

  • If you use encrypted communications with your SQL Server, have you noticed a drop in performance (from not using encrypted communications)?

    I am running it now, and I keep hearing or reading that it will affect performance negatively. I decided to do my own simple test. If anyone can point out a problem with my test, please do so, as I want to really know how much encryption affects performance.

    My test:

    The test was on a Dell 6850, dual 3.0Ghz with 8G of ram, 7G dedicated to SQL Server, and consisted of running the same query over and over, restarting SQL Server after changing the Force Protocol Encryption option between each run. This means SQL Server has no cache built up, that could affect the test. The results were the time period Query Analyzer said it took to execute the query.

    The query I ran would normally return over 100K records, but for this test I limited the result set by using the Set Rowcount command, setting the rows returned to 20,000.

    Do you think this test is valid? If not, what do you suggest for a valid test?

    Here are the results:

    0:44 no SSL

    0:32 SSL

    0:34 no SSL

    0:31 SSL

    The SSL queries were faster than non-SSL. How can this be?

    Thanks,

    Chris

  • This was removed by the editor as SPAM

  • In running some tests a few years ago, I did notice some degradation but never took it a step to actually measure how much. I don't think it was significant, but then again, I wasn't running any significant processes, either.

    It's hard to say if your test is valid because you've only talked about the server itself and the mechanism of connection. Was the environment isolated to a client, a DC, and a SQL Server? That way extraneous network traffic isn't an issue, etc. Also, I'm looking at the range in times between the no SSL and it's 0:10. That would seem to indicate other things are going on...

    K. Brian Kelley
    @kbriankelley

  • I have not really thought of a more valid test for this, if this one is invalid. My testing had no isolation or segmenting of networks. My main concern was how much more load would be on the SQL Server to do the encryption when returning results.

    We are putting a hardware product in between the SQL Server and network today to measure network traffic for both scenarios to see how much overhead SSL causes.

    Thanks,

    Chris

Viewing 4 posts - 1 through 3 (of 3 total)

You must be logged in to reply to this topic. Login to reply