Router Throughput

[dhellwig]

Does anyone know what kind of throughput can be expected from various router platforms? Customer has some high-end internet service with CPE dropping off FastEthernet. Wants to know if he sticks a 2651XM with two fastethernet between that and the outside of his firewall if the router will choke his internet bandwidth. The router will also have some WAN links to an alternate site and run BGP for redundancy to the internet.

Is there any argument for using a 2691 or a 3745?

Thanks,

...

[tbaranski]

Information on router throughput capability is hard to come by because maximum acheivable throughput depends greatly on average packet size (the packets-per-second capability of a given router tends to decrease as packet size increases, but throughput tends to increase because the packets are bigger) and enabled features (functions such as ACLs and NAT tend to reduce performance).

So whether or not a 2651 will be a bottleneck on your customer's network depends on the throughput the router will have to acheive (Internet bandwidth + WAN links), packet size, and enabled features.

Some quick ballpark figures to use as a reference: under optimal conditions (full-sized packets, CEF enabled, no performance-reducing features enabled), a 2621 can do around 40Mbps, a 3620 can do around 75Mbps, and a 3725 can do over 200Mbps. But keep in mind that real-world traffic flows are never optimal.

[andyirving]

How did you calculate these figures, I too am trying to get some ballpark figures on what a 2600 series router can throughput. Take a 2621 with dual Fast Ethernet. Its stated performance is 25kpps.

At 64 byte packets this works out as 12.8Mbps.

25000 x 64 x 8 = 12800000. Does this mean if I have 100% bandwidth on one of the Ethernet ports it could throughput that at wire rate onto the other Ethernet port?

[tbaranski]

The above figures weren't calculated -- they're based on performance tests that have been done by myself (2621 and 3725) and others (3620) with full-sized (1500 bytes) packets.

Cisco's pps ratings are really only good for estimating relative performance; i.e., comparing one router's pps rating to another router's pps rating. The actual throughput potential depends on many factors as I mentioned above.

A 2621 will never do 100Mbps wire speed. But I only know this for sure because I've tested one. But based on the numbers I gave above, the maximum throughput for a given router based on Cisco's pps ratings (pps * 64 bytes) seems to increase by 3-4x as packet size increases to 1500 bytes. The 2621, for example, can do 12.8Mbps with 64-byte packets per Cisco's pps rating, but can do ~40Mbps with 1500 byte packets per my testing. But this is just a very rough way of ballparking a router's potential throughput under ideal conditions.