What is the meaning of Forwarding capacity or Switching capacity of a switch ?

[Soppimath]

What is the meaning of Forwarding capacity or Switching capacity of a switch ? what we should need to understand when we read Forwarding capacity up to 720 Mpps ?

Example :-

Cisco Catalyst 6513-E Switch:-

13 slots, 19 RU
44 x 40 GE ports
180 x 10 GE ports
534 x 1 GE ports
529 x 10/100/1000 ports
Forwarding capacity up to 720 Mpps <<---- what is the actual meaning of this ?

Cisco Catalyst 6880-X Switch:-

4 module slots, 5 RU
80 x 10/1 G SFP
Slot capacity 80 – 240 Gb/s
Switching capacity up to 4 Tb/s (VSS)<<---- what is the actual meaning of this ?
Cisco Catalyst Instant Access

[Joseph W. Doherty]

Both mean what they say.

"Forwarding capacity up to 720 Mpps" means the 6513 can forward up to 720 million packets per second. (NB: with PPS, size of packets can impact PPS rate. A 6500's PPS rate may be packet size independent.)

To obtain such PPS rates on a 6500, you need to have DFC equiped line cards.

Also BTW, 1 Gbps of minimum sized Ethernet packets is 1.488 Mpps.

"Switching capacity up to 4 Tb/s (VSS)" means a pair of 6880s can forward, across their fabrics, an aggregate of 4 terabits a second (each 6880 has a capacity of 2 Tbps - the 6880 uses a sup2T).

[Soppimath]

Thanks a lot Joseph, 

Could you please let me know in general how we can understand switching capacity of a switch ?

Whether it depends only with it's supervisor ports alone or number of line cards ?

whether switching capacity is "forwarding capacity of the supervisor+number of line cards" in the entire particular chassis ?

Let us take we have one small switch with number of ports only 10 , now let us take it's forwarding capacity is 100 packets per second .

take the ports are 1, 2, 3, 4, 5, 6, 7, 8, 9 and 10.

means when we connect transmitting host at port 1 and receiving host at port 2 , then do we need to understand, the switch can forward max 100PPS ?

or  transmitting host at port 1 and receiving host at port 2, 3, 4, 5, 6, 7, 8, 9 and 10 , then do we need to understand, the switch can forward max 100PPS ?

[Joseph W. Doherty]

"Could you please let me know in general how we can understand switching capacity of a switch ?"

Often an internal aggregate bandwidth capacity is noted along with an aggregate PPS capacity. The internal aggregate bandwidth capacity would need to be at least the sum of all the port bandwidths (twice that if ports are full duplex). (Some switches do not have such capacity, and if not, they can bottleneck port to port transmission. Some switches note higher than what the ports could use. This might be to support add-on modules or it might just be additional capacity not being used. An example of the latter, might be an switch fabric designed for a 48 port switch being used for only 24 port model.)

PPS capacity would be what's available for forwarding packets, often as an aggregate. PPS needs vary per frame/packet size. (Smaller needs a higher PPS to utilize the same bandwidth. Some switches cannot sustain wire rate with small frames/packets.)

For an example of a 48 GB (duplex) switch, that supports wire rate, its specs would generally list its fabric capacity as 96 GB and its minimum size Ethernet frame rate as (about) 72 Mpps (1.488 Mpps per GB).

"Whether it depends only with it's supervisor ports alone or number of line cards ?"

Chassis can become rather complex. For example, understanding a 6500, generally requires a multi-page write up. In the 6500 series, there are differences between chassis models, between supervisors and between line cards (and how all the forgoing interact) that impact the chassis's bandwidth and PPS rates.

"Let us take we have one small switch with number of ports only 10 , now let us take it's forwarding capacity is 100 packets per second .



take the ports are 1, 2, 3, 4, 5, 6, 7, 8, 9 and 10.



means when we connect transmitting host at port 1 and receiving host at port 2 , then do we need to understand, the switch can forward max 100PPS ?



or transmitting host at port 1 and receiving host at port 2, 3, 4, 5, 6, 7, 8, 9 and 10 , then do we need to understand, the switch can forward max 100PPS ?"

In your example, your two performance bottlenecks would be the bandwidth capacity of port 1 (assuming it and all the other ports have the same bandwidth capacity) and/or the aggregate 100 PPS. As such, unlikely it would matter whether egress port was just port 2 or ports 2..10. (NB: I write unlikely, because there are switches whose behavior is controlled by architectural features more than aggregate bandwidth and/or aggregate PPS. For example, whether two ports are on the same ASIC or module can be important to performance.)