09-08-2014 11:41 AM - edited 03-07-2019 08:40 PM
I am trying to understand how to connect 5 switches using 0.5 M cables only. I have read this "best practices" document, but it leaves some unanswered questions.
I would normally just daisy chain the switches and bring a connection from the bottom switch back up to the top to create a "ring" topology. But the documentation also states that switches can be connected with "interweaved cables". It makes no statement as to what if any performance considerations should be taken into account when designing stack interconnections, or whether this interwoven topology creates any operational complications. The only requirements that I have seen is that each switch must have 2 operational FlexStack links in order to be fully redundant and to fully utilize all available bandwidth. Does anyone have any guidance in this area?
Assuming no negative consequences by using the "interwoven" topology, does this make sense?
Thanks,
-Jeff
09-08-2014 03:19 PM
It'll work.
Someone has decided to "re-invent" the wheel and made the cabling confusing instead of the KISS principle.
09-08-2014 10:00 PM
Hi Leo,
I have in production 1x WS-C2960S-48LPS-L 15.0(2)SE2 C2960S-UNIVERSALK9-M
and want to stack a new one switch (WS-C2960X-48LPS-L 15.0(2)EX3 C2960X-UNIVERSALK9-M) to it.
Would it work like this or do I need to upgrade IOS or downgrade IOS?
Please advise
Thanks
Ql
09-09-2014 02:58 AM
QI,
You will need to "downgrade" the IOS of the 2960S to the same level as the 2960X. I would not recommend anyone use 15.2(2)E.
Take note that you are "cross" stacking a 2960X with a 2960S. This means that you can stack a maximum of four 2960S/2960X.
09-09-2014 06:29 AM
Well it probably wouldn't be confusing if I could find a clear set of ground rules. The documentation mentions "FlexStack protocol", but the discussions seem to indicate that this is a control plane protocol. Do the stacking modules operate as ethernet bridges when making forwarding decisions like any other switch port? If the answer is yes then at least I understand the principal of operation. But now if you throw multi-chassis etherchannel uplinks into the mix divining general layer 2 paths becomes a little murky.
Although now that I look at it, using this cabling has an advantage over a straight ring topology. Once a frame hits the stack it is never more than 2 hops away from it's destination. With daisy-chained switches it could be up to 4 hops. This is of course assuming standard 802.1d-like forwarding. This may be an invalid assumption. I don't see any way to get spanning tree info for the stack ports.
(EDIT: standard ring topology would also have a maximum of 2 hops to any switch, not 4)
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