Reply
Contributor
taglia
Posts: 11
Registered: 06-10-2009
0
Accepted Solution

Strange performance impact of WDE

I have been using PGP WDE for some weeks, without any visible impact on performance. Having two almost identical machines at hand (same model, CPU, memory, and disk; one fully encrypted, the other with no WDE), I tried to measure this.

 

To my surprise, QuickBench reported a performance hit of roughly 35%. You can check the results on this post. To be honest I am quite puzzled by this, as 35% should be very visible on the normal work. I did a quick test with XBench as well, with similar results. Suspecting some other applications being responsible for the performance degradation, I booted from an external disk with a vanilla Leopard install, plus WDE: I used this disk just to boot, and tested the internal encrypted disk. Again, same figures.

 

Does anybody have a rational explanation? With the current data, I can only conclude that disk benchmarks are not a good performance measure, as they do not reflect typical disk usage.

PGP
duane
Posts: 572
Registered: 01-15-2006

Re: Strange performance impact of WDE

Your conclusion is fundamentally correct. XBench is designed to test disk performance. As such it bypasses the internal file system level buffers that provide much of the performance improvements that you see when using your Macintosh. 

 

XBench is a good measure of raw disk performance, but not of typical file system usage.

Super Contributor
lhotka
Posts: 758
Registered: 01-26-2007

Re: Strange performance impact of WDE

I've confirmed this with real-world benchmarks.  The actual hit for normal use is about 5%.  I only see bigger hits if I'm copying very large files on a high-speed bus, with high-speed drives.  I suspect that what happens is that the benchmarks saturate the WDE buffers, whereas real-world use rarely does.