VPN Performance Study

[dnewman at cmp.com]

Thanks very much for this pointer. Good stuff indeed--I'd yet to see any latency
numbers, and the comparisons of various algorithms are fascinating.

A couple of follow-up questions for Tom:

1. Have you performed any tests with newer Linux kernels? The IP stack in the
1.X.X kernels was significantly slower than the 2.X.X stuff (<religion> or than
most versions of *BSD, for that matter </religion>), and that may be a factor
when using a tool like TTCP.

2. Have you tried a UDP-based traffic generator and very short packets, or some
mix of long and short packets? TCP and relatively long packets alone give the
devices a fair amount of breathing room.

Again, thanks for this excellent resource.

Regards,
David Newman
Data Communications magazine






Tina Bird <tbird at secnetgroup.com> on 09/09/99 12:41:52 AM