Hi, I know this is a trivial question of sorts, but wanted some better feedback.
Lets say I am working with a 4011 or CCR1009 copper router. I have a 3 managed switches, 2 in the main rack and one switch in a remote rack. For sake of discussion we can pretend they are 48 port switches. The devices are data drops for offices, VoIP phones, IP cameras, access points and some IoT devices. Perhaps a small NAS or server for fire sharing in the main rack. VLANs are being used.
Would I be better off creating 3 trunk ports on the router to connect each switch in a “star” topography? Or should I daisy chain the switches together? I am having a hard time figuring out why daisy chaining switches seems to be the standard. I know you can have some redundancy if you loop them together, but you eliminate this if you don’t chain them together in the first place. Considering many switches some kind of 10G ports that you could link together use to connect them and achieve higher bandwidth between them. Would connecting them all to the router be a potential bottleneck?
I suppose if you were talking about 50 switches you would have not choice but to daisy chain them, so perhaps that scales better. It seems like bandwidth between the switches in a smaller system is the most compelling reason to daisy chain - is that right? How would the design change if something like the CCR2004 was used?
Thanks in advance for any help. I just want to make sure I am doing this the right way.
The interconnection topology should respect the traffic topology at first place. If there is no significant traffic volume between two switches, then there is no need for a direct interconnection between them, so the star topology makes more sense, and even for the backup links it is better that they form up a parallel star. When talking about more even traffic patterns, a mesh interconnection topology gives you more bandwidth (but mere xSTP doesn’t automatically discover the shortest path through the network, you need a mesh protocol or MSTP and manual configuration of STP instances handling different VLAN groups, or you need routing with a dynamic routing protocol). And if bandwidth is not your concern, a simple ring gives you a better protection against mechanical damage, but it needs a careful design to make real sense - I’ve seen cases where the excavator dug exactly where both cables were leaving the building through the same duct so the site got cut off the rest of the network despite the logical topology was a ring one.