2011-09-01 14:24:32

by Adam Tisovsky

[permalink] [raw]
Subject: IPsec performance (in)dependent on ingress rate?

Hi,

I?m doing some benchmarks of IPsec performance on Cisco router and I
have experienced the situation described below. My question is whether
anybody has performed similar tests on Linux (StrongSWAN, OpenSWAN,?)
or any other security gateway and can tell how did it behave.

When you are gradually increasing the rate of traffic to be secured
(using UDP as a transport protocol) you reach the maximum possible
throughput of the device. But when you continue increasing the rate of
ingress traffic beyond this point, the fowarding rate of device will
decrease. Example:

Max. throughput of device is 10 Mbps. If ingress traffic rate is 10
Mbps, then forwarding rate is 10 Mbps. But when ingress rate is 20
Mbps, you get forwarding rate only 5 Mbps.

I have experienced this on Cisco 1841 router with HW accelerator
DISABLED. After some investigation I foud out that more ingress
traffic utilizes main CPU more by interrupts. And interrupts go on the
expense of encryption process. Therefore the decrease of forwarding
rate. With HW accelerator enabled this situation on does not occur,
device forwards traffic at the maximum rate even if it?s overloaded by
the ingress tarffic.

I didin?t find any information dealing with this, however I find it
quite interesting. I?m also planning to do the tests on StrongSWAN and
OpenSWAN, but it takes some time. So any information will be helpful
in advance.

Thank you
Adam


2011-09-05 02:26:03

by Sandy Harris

[permalink] [raw]
Subject: Re: IPsec performance (in)dependent on ingress rate?

On Thu, Sep 1, 2011 at 10:24 PM, Adam Tisovsky <[email protected]> wrote:

> I’m doing some benchmarks of IPsec performance on Cisco router and I
> have experienced the situation described below. My question is whether
> anybody has performed similar tests on Linux (StrongSWAN, OpenSWAN,…)
> or any other security gateway and can tell how did it behave.

There is some info for FreeS/WAN, ancestor of the ones you mention:
http://www.freeswan.org/freeswan_trees/freeswan-2.06/doc/performance.html