I'm installing a stack of 3x 3750G switches, which are inter-connected w/ StackWise cables.
Are stacked switches truely act as one switch w/o any performance penalty if a packet has to traverse from one switch to another in the same stack?
Would it be better to arrange the ports together on the same switch if I know they'll mostly generate local traffice among themselves?
Another question I have is sould I try to distribute ports among the stacked switches as much as possible or is the stack really functioning as a unified stack??
For instance, I've got 3x 24-port 3750G's; thus a total of 72 ports. If I currently have only 24 objects (clients or switches), should I plug 8 in each switch or it doesn't make a difference if I plug all 24 objects into the same switch?
With Cisco's stckwise technology, its a true stacking feature with makes your all the switches in a stack as a single switch. All the switches connected via stackwise cables have the common backplane of 32 Gbps and really act as a single switch. The whole stack is managed by a single IP.
This also gives you a n:1 master redundency, which means that when the active master fails, the other switch immediately takes over as a master switch.It really doesnot make any difference of plugging all the ports in a single switch.
its not that only if you connect 9 switches you get 32Gbps of backplane, even if you connect 2 or 3 switches stacked together also you get 32Gbps of back plane. even a single 3750 switch back plane capacity is 32Gbps.
having the switches stacked together in 3750, it assumes as one single switch virtually & physically 9 or 2 or how many switches you stack.
A customer asked me this exact question the other day....
The stack has an interconnect speed of 32G. Thus a packet coming in from 1/0/1 going to 3/0/1 will hit the backplane on switch 1 (speed so high you can forget about it for all intents and purposes) traverse the 32G interconnect bus, across the backplane on switch 3 to the egress port. You can consider this almost identical to the 6500 series when using the "classic" series line cards (61xx series) which likewise share a 32G backplane bus.
I have tested this with IXIA gigabit traffic generators which blasted 20 Gig of traffic with 50000 flows across the switch stack. The boxes were solid as a rock. 20Gig is rather a lot of traffic.... !!!
The only occassion when you might have a problem with this architecture is if you have multi gigabit traffic crossing the switch stack that requires QOS priority on an already heavily utilised stack. In that case you could possibly get contention for the stacking bus (unless there is a way of prioritising the traffic gaining access to the stacking bus???? There hasn't been in the past). Cisco's answer to that would be twofold.
1.If you required such traffic patterns with such heavy traffic and QOS requirements then you should be using a 6500 series switch with high end interface blades.
2.That the 3750 supports ingress QOS to precisely address the problem of overloading the switch fabric.
We are pleased to announce availability of Beta software for 16.6.3. 16.6.3 will be the second rebuild on the 16.6 release train targeted towards Catalyst 9500/9400/9300/3850/3650 switching platforms. We are looking for early feedback from custome...