I have 3 Cisco switches connected daisy chain ?

[SEBassman]

I have 3 Cisco switches connected daisy chain one to another to another. I have read that I can and I can not do this so I'll ask. Is there and advantage in using 2 patch cables between each switch? I've read it is suppose to increase performance. but also seen that it creates routing issues. SO is this a good or bad ting to do?

[balaji.bandi]

STP will block one link to prevent bridging loops when you connected them in RING.
You can play with STP cost and priority different vlan to prevent this.
Rapid PVST is suggested here

Note : why not you make a Stack if they all in one location, what is the use case here

[Seb Rupik]

Hi there,

As @balaji.bandi mentions, when using STP any forwarding loops will be prevented, under normal conditions.

With regard to using two cables between each switch, what you are elluding to is the configuration of Etherchannels. This will provide a logical interface with the aggregate bandwidth of the member interfaces as both will be in a forwarding state. Keep in mind that Etherchannel traffic is load-balanced per-stream not per-packet, so under some conditions a number of the member links may not be utilised.

cheers,

Seb.

[Joseph W. Doherty]

So, you have something like sw1<>sw2<>sw3?

If so, sure you can do that.

Is doing this "bad"?  Well that's an "it depends answer. If most of your traffic is between switches 1 and 3, it needs to transit switch 2.  (BTW, you could also look as this 3 switch topology as a "star" or "tree" topology centered on switch 2.)

For redundancy, if possible, you might also interconnect switches 1 and 3, creating a "ring", although, as noted by @balaji.bandi, (if doing this at L2) you'll need something (e.g. STP variant) to "block" one of the links, between switches.

If your switches are L3, in such a ring configuration (also in this instance a full mesh), L3 can use all links, concurrently, and chose the "best" path between switches.

[SEBassman]

Thank you all for the replies. Sorry for not providing more info. Gigabit Comcast service Modem/router > Switch 1 > Switch 2 > Switch 3.

Switch 3 is a (Cisco 2960 - 48) Poe Switch for the phones. 1 & 2 (Cisco SG200 - 26) for the PC's & printers portion of the network. Looking to get the best speed and performance out of this simple configuration. Was not sure if 2 cables would make any difference performance wise.

thanks for the help.

[balaji.bandi]

Daisy chain or Ring  will not give you more bandwidth, it gives more other issues if you do not configure it properly

I suggest below design to make it proper. rather dais chain.

Note: thinking that all in one place.

[Joseph W. Doherty]

"Was not sure if 2 cables would make any difference performance wise."

It's an "it depends" answer.

If you're passing much data between switches, i.e. enough to saturate your single link, then something like a dual Etherchannel link (if configured optimally), generally would provide about 50% more bandwidth.

BTW, adding a second link, non-Etherchannel, will not help (much [NB: as Cisco's STP is per VLAN, it's possible to "map" VLANs to link usage, i.e. allowing usage of a link that non-VLAN STP would block]) as STP (which should be enabled) will block links such that's there's only one active path.

And/or, as @balaji.bandi's post, i.e. the one with diagram, can help too as it eliminates switch2 as a transit for switches 1 and 3 or router and switch 3.

To recap, if, and only if, your existing bandwidth is limiting performance, you might use Etherchannel and/or change the topology to "star".

[Richard Burts]

I would suggest that there are really 2 parts to this question: daisy chain switches and more than one cable connecting switches. Here are my responses:

1) While daisy chain connections are easy I do not think that they are a good thing to do. My primary reason is that if there is some issue on SW1 then SW2 and SW3 may lose their connectivity, and some issue on SW2 could impact SW3. I believe that the suggestion by @balaji.bandi is better than a daisy chain.

2) If switches are connected by a single cable then you have a significant single point of failure. If there is a problem with a cable then any downstream switches are dead in the water. Connection with 2 cables (2 cables between access ports for a single vlan or 2 cables between trunk ports for multiple vlans) will provide redundancy. As noted in previous responses 2 cables between switches presents a loop and Spanning Tree will put one cable into blocking. In case of a failure of the active cable then the second cable becomes active. But there is no change in performance. If the 2 cables are configured to create an EtherChannel then you get the advantage of redundancy and possibly get an improvement in performance.

Having explained this logic I now realize that I have been making an assumption (and suspect that other responders have also) that we are talking about layer 2 switches. But perhaps these switches are operating as layer 3 switches? If they are layer 3 switches then I believe that there is an even stronger case that 2 cables would be a "good" thing. With L3 switches connected by 2 cables then you could run some dynamic routing protocol between switches. The dynamic protocol would track the status of the cable - can we maintain a neighbor relationship over this cable - if so then use the cable, and if not maintain a neighbor relationship stop using the cable and automatically fail over to the active cable. So with L3 you would get dynamic failover and the performance advantage of equal cost load sharing.

[Joseph W. Doherty]

@Richard Burts's post "pushes", I believe, redundancy.  I very much like redundancy too.

Usually with a small switch, I worry more about switch failure than cable or port failure.  I.e. As Rick, and I earlier, pointed out, with a "daisy chain", failure of a transit switch can disconnect other switches (and their connected hosts).

However, @balaji.bandi's configuration, except for moving the "router" the "core" switch, is still, for the switches, physically a daisy chain.  Adding dual links, between devices in Balaji's configuration still leaves the "core" switch as a critical failure component.

What can be done to further improve redundancy (at minimum cost)?

First, again in Balaji's topology, also (assuming its possible) connect the two SG switches; creating a "ring".  You'll want, in STP, to make the "core" switch (the 2960) the "root" STP switch.

Further, as this new link is your "compact spare tire", using an "ordinary" edge port can be okay.  If fact, as often switches support "banks" of ports on the same internal hardware, ideally you want redundant ports as far apart as possible, like the first and last edge port, or an uplink port and an edge port.

Again, for the links that Balaji has, between switches (still assuming L2), Etherchannel (if supported) might be used on those links.

With the above mentioned additional link, creating a "ring", the "core" switch is still critical to getting to/from your router.  So, ideally, we would like a second link to the router (again, might be just a "compact spare tire") connecting to another switch.

If one or more switches can do L3, you can "easily" configure two separate router<>switch links.

If switches are only L2, the router could connect to two different switches, if it also supports bridging. (Rarely used, so might be more difficult to configure.)

Anyway, to recall your original questions, can you do daisy chain?  Yes you can.  Should you?  If possible, best to avoid (as it often adds traffic to links that might otherwise need not be there; also can add latency, to transit other devices).

Can multiple links be added between devices?  Depends on the device capabilities, and whether connections will be L2 or L3.  Will additional links improve performance?  Maybe yes, maybe no.  (BTW, generally for both L2 and L3 multiple links, link loading isn't considered.  I.e. new traffic might be directed to a saturated link while another link is totally unused.)