07-21-2017 07:25 AM - edited 03-08-2019 11:25 AM
Hello,
I have a quick question. I have two old Catalyst 3750s set up in a stack using the correct ring setup. The switches have 100Mb/s ports on them. They are only connected via stacking cables. I have two physical servers, one on each switch. When I transfer a file between the two physical servers, the data transfer is at 100Mb/s. Is this the correct and expected behavior? I can't find much documentation on how Stackwise works, logically.
Thank you very much, all!
Solved! Go to Solution.
07-21-2017 08:03 AM
So the switch's physical port is still causing the bottleneck, though?
Correct.
The StackWise show commands state it is in a 32Gbps duplex configuration.
Yes it would, but like fabrics, Cisco counts the 8 Gbps duplex as 16 Gbps, and as you have two, you have "32 Gbps". But from one stack member to another, you wouldn't be able to transfer unidirectionally at more than 16 Gbps.
07-21-2017 07:34 AM
Hi,
That is the correct behavior. The NICs on the server are 100Mb and that is the maximum transfer speed you get.
HTH
07-21-2017 07:41 AM
As Reza noted, you're bottle neck will be the server FE connections.
But, if you're wondering, original StackWise I believe, is 8 Gbps, duplex, per stack port. StackWisePlus is double that and for unicast traffic, uses the stack ring much "better".
07-21-2017 07:53 AM
The NICs on the servers are 1Gb/s, whereas the ports on the switch are 100Mb/s. The StackWise show commands state it is in a 32Gbps duplex configuration.
So the switch's physical port is still causing the bottleneck, though?
Thanks!
07-21-2017 08:03 AM
So the switch's physical port is still causing the bottleneck, though?
Correct.
The StackWise show commands state it is in a 32Gbps duplex configuration.
Yes it would, but like fabrics, Cisco counts the 8 Gbps duplex as 16 Gbps, and as you have two, you have "32 Gbps". But from one stack member to another, you wouldn't be able to transfer unidirectionally at more than 16 Gbps.
07-21-2017 08:07 AM
Gotcha! Okay, I understand. So, the 32Gbps backplane is just to pass traffic from the up-to 9 stackable access switches between one another and then up to a distribution switch (or two) via an uplink port on each switch for routing if need be?
07-21-2017 08:23 AM
Yea, but again, there's a big difference between how StackWise and StackWisePlus work.
With StackWise, every switch stack member places all its internal traffic on the ring, and the stack members that put the traffic on the ring also removes it. Effectively, StackWise performs like a shared bus.
With StackWisePlus, a stack member only places unicast traffic on the stack ring if its destination is another stack member and the destination stack member removes the traffic from the ring. Effectively, the stack behaves as a stack of cascaded switches (again, at least for unicast traffic).
07-21-2017 08:27 AM
Oh, that's pretty cool! I'll keep that in mind.
Thanks for the help and sharing the knowledge. I really appreciate it! Have a good one!
- Tyler
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide