Switching

i want to understand term of non blocking

i understood it means that switch can handel all traffic coming on all ports ( maximum speed ) with no delay 

so for example , for EX4200 48 Port  ... if i want to prove that it is non blocking , how can i do it by calculations ? is it like :

48* 1G  + 2*10 G = 68 Gbps ( so backplane should be at least  68 Gbps ) ?

Don't forget full duplex -- this is a switch.  

[(48*1G + 2*10G) * 2] = 136Gbps

All ports can simulaneously forward and receive traffic at full line rate.

Hi @Keithr,

So do we have switches within the Juniper portfolio that are NOT "non-blocking"? And what would it be correct if I say the opposite of non-blocking is "over-subscription":?

---

@wendohw wrote:

 

So do we have switches within the Juniper portfolio that are NOT "non-blocking"? And what would it be correct if I say the opposite of non-blocking is "over-subscription":?

---

I really don't think I'm qualified to fully answer that question...

"non-blocking" can have different answers depending on the specifics of the question being asked.  Non-blocking at full-sized frames (1518 bytes) might not always be non-blocking at 64 byte frames.  You'd really be best served to speak with a SE.  Make sure you ask specific questions, though, because SE's are sales people in disguise... they'll often answer enough of your question to be able to say "yes" and not volunteer the rest of the story unless you pull it out of them.

The opposite of non-blocking is not necessarily oversubscription, but in the context of within a single switch, then sure, you could say that.  If a switch is fully non-blocking, that means that its port-to-port switching should never be oversubscribed.  Once you leave the switch (hit an uplink, for example) all bets are off.  You could have a fully non-blocking switch, and if you have 20 hosts sending full rate traffic on 20 1G ports and that traffic has to exit the switch via a single 10G uplink, then that *link* is oversubscribed, but the switch itself isn't.