Each packet that travels through the 2950 internally has headers added reflecting the output ports, vlan, prioritization, and so on. Those internal headers take up internal bandwidth.
The switching fabric rate reflects the internal bandwidth. In systems with multiple line cards, the fabric rate deals with how quickly the (internally augmented) data can be shipped around. The forwarding rate deals with how quickly (aggregate total) the device can pull data onto and off of the wires, and takes into account internal queue sizes, internal classification and prioritization and routing or time to decide upon the destination ports.
If the switching fabric rate is not comfortably more than the forwarding rate, then it implies that there is are "choke points", traffic paths that cannot operate at full speed.
A good example of this is the 6500 series architecture, which has gone through several generations of backplane connections for the peripherals, some of which were zippy in their day but now pretty much crawl, relatively speaking. It is not uncommon on the 6500 to find line cards that can switch or route very quickly to other ports on the same card, but because the switching fabric is relatively limited, routing or switching to other cards might be much more limited. You can get line cards for the 6500 which have 48 gigabit ports, but which can only attach to the backplane (switching fabric) at
6 Gb/s, so you if you did not know the details of the buffering and controller arrangements, you could end up "oversubscribed" by 8:1 if groups of ports happened to need the switching fabric simultaneously -- the switching fabric would "choke" the performance down to 6 Gb/s when travelling off-card.
(Note that this is not necessarily a "massive design flaw": if those
48 ports are connected to desktops, chances are excellent that only a few of them are actually trying to use the switch for any kind of sustained gigabit transfer.)
Let's see...
OM2: 50 um or 62.5 um multimode, 500 MHz* km modal bandwidth at
850 nm or 1300 nm
Glancing at a draft for 10000Base-SX it appears the above would be within spec for the following combinations:
50 um MMF, 500 MHz*km modal @ 850 nm, 2 to 86 metres
62.5 um MMF, 200 MHz*km modal @ 850 nm, 2 to 35 metres.
Glancing at a draft for 10000Base-LX, it appears the above would be within spec for the following conditions:
50 um MMF, 500 MHz*km modal @ 1300 nm, 2 to 300 metres
62.5 um MMF, 500 MHz*km modal @ 1300 nm, 2-300 meters
You should crosscheck that those made it into the final specs.
Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.