[c-nsp] Cisco 3640 Maximum performance

[Rick Havern]

>From lab testing I've done, it appears that the 3640 was designed to
terminate a T3 and nothing more.  Using CEF and no fancy packet processing,
45M is when the CPU starts to melt.

Rick

####

>From Alex Rubenstein

Nothing personal, but you're insane. The 3640 has a hard enough time
moving 50 megabits/sec in the so-called 'real-world', even with CEF. I
recall using a 3640 with a FE and a NM-HSSI to terminate a Verizon DS3
Frame Relay port with a few hundred PVCs we once had. When the port would
see anything near 30 to 40 mb/s, CPU would creep up into that threshold of
'DANGER WILL ROBINSON' -- about 60%.




On Thu, 30 Sep 2004, Koen Peetermans wrote:

> Actually, you should get a lot more than 70-90Mbit/s out of the platform.
>
> Let's assume a forwarding rate with 50.000 packets per seconds and packets
> of 1000 BYTES, That gives 50.000.000 BYTES/Second = 50 MBYTE / Second =
> +- 500 Mbits / Second.
>
> The actual performance can be higher or lower, depending on
> software/hardware features used.
>
> Kind regards,
>
> Koen.
>

<snip>