[c-nsp] new Cisco routers 1800/2800/3800

[Church, Chuck]

That 115200 figure came from 225,000 pps x 8 bits/byte x 64 byte packets.  That performance you mentioned for a 3745 was using the same size packets that Cisco tests with.  I don't think you'll see 200 kpps with real-world  packet sizes and thousands of simultaneous flows.  100 kpps under these circumstances is probably more accurate.  Also, pps will vary with packet size.  If you make the change to testing with 640 byte packets instead of 64, you won't see 10x the throughput, at least not on a software-based router.  An ASIC based switch probably would scale appropriately.  I believe the PCI bus, DRAM speed, and other factors limit this.  Like you said though, I bet with 64 byte packets that 3845 will exceed 300 kpps.   Overall, I bet these 3800s and 2800 will be about a 2x improvement over the 3700s and 2600s they're replacing.  Looks like competition from Juniper is forcing Cisco to put some pretty capable routers out.


Chuck Church
Lead Design Engineer
CCIE #8776, MCNE, MCSE
Netco Government Services - Design & Implementation
1210 N. Parker Rd.
Greenville, SC 29609
Home office: 864-335-9473
Cell: 703-819-3495
cchurch at netcogov.com  <-note new address!
PGP key: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x4371A48D


-----Original Message-----
From: cisco-nsp-bounces at puck.nether.net [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of Lukasz Bromirski
Sent: Thursday, September 30, 2004 10:56 AM
To: cisco-nsp at puck.nether.net
Subject: Re: [c-nsp] new Cisco routers 1800/2800/3800

Church, Chuck wrote:

 > Anyway, I think we've found Cisco's rating of pps and throughput to be
 > based on 64 byte packets and no interface config other than IP
 > addresses.

Yes, but some info is based on packets with 256-bytes or 512-bytes.
I don't have any URLs at hand, but some presentations and whitepapers
mention that.

> Cisco lists the 3745 at 225,000 pps and 115,200 kbps.

115,200kbps....of what? I don't see any figure close to this, but
only max console/AUX port speed...

 > I don't believe anyone here has seen this router even get close to
 > half that number of pps.

Then believe. I've got 3745 under tests with and without BGP, it
was doing close to 200kpps with Fastethernet<>Fastethernet routing
(no services/filtering/NAT). I.e. it was routing full 100Mbit/s of
64-bytes with CPU reaching 85-90%. It could do more, if the
packets weren't so small, i.e. it would fill the pipe much faster.

 > This new 3845 did about 95,000 pps at full size and with CBAC and
> NAT. 

...and it's certainly not the best pps figure You will find in
marketing materials. Good looking figures are high ones, and usually
pps values are given for 64-bytes packets, as they will be higher.
If 3845 can route 95kpps for 1500-bytes packets, it will route
more for 64-bytes - that's just math. Of course, the more the
packets in any slice of time, CPU/interface ASICs will do more
work, but this figure is very low.

-- 
this space was intentionally left blank    |              Łukasz Bromirski
you can insert your favourite quote here   |          lukasz:bromirski,net
_______________________________________________
cisco-nsp mailing list  cisco-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/