How does SSL Impact Server Performance

Hello,

I run web-based survey software on my own server (A Dell PowerEdge 600SC with 3GHz P4 processor, 2 GB of RAM, SCSI drives, Windows Server 2000 and IIS 5.0). Some of the surveys I host collect sensitive information (not credit card data but other information respondents view as confidential). I am thinking of adding SSL to my website to give respondents an extra sense of peace of mind and security when they complete my surveys.

I have heard that SSL adds overhead to all transactions due to the multi-step handshake process and the time involved in generating the session keys.

My question is how much, if at all, will SSL impact my server's performance? The number of active sessions flunctuates quite a bit. 98% of the time I have 100 or less. I occassionally get as high as 200-300. My pipe is business-class 768 SDSL. To date without SSL I have never head a performance bottleneck.

Does anyone have any insights or suggestions as to how I can estimate the impact of SSL on my server's performance?

Thanks so much!

 

Mr Chip,

To get the exact figures on how SSL will impact your web site you can use performance testing tools to gather benchmarks before and after implementing SSL.

Mercury Load Runner is a good product to use for performance testing but it comes with a hefty price tag. If your working with a small budget I recommend using open source tools to test performance.

Reference: http://opensourcetesting.org/performance.php

-Freedom

 

Arie,

Thank you for the valuable insights! Based on this information, I have changed my plans. I do not want to burden all of my users with SSL (since only 10% will need it). I am looking into how I can easily implement 2 sites. One with SSL and one without.

As an aside, I am sorry I did not respond sooner. Twice I tried to post a reply, but ran into an error.

 

Thank you for the additional information. I will look into the open source option for my SSL site. Sorry for the slow reply - I ran into technical problems when I tried to reply previously.