3850/3750 Standalone or Stacked Distribution Layer Question

[networkwise]

Hello All,

I have a customer with three buildings all very close to each other MM 62.5  fiber backbone. They've outgrown there current topology and I'm looking at implementing a distribution layer using a couple of either 3750's or 3850's. They've already invested in Netgear at the access layer and I'm attempting to make the two work together Cisco and Netgear. My experience is with Cisco don't know Netgear that well. The mixed Vendor environment needs testing, the thing Im not sure of is using a stacked distribution layer with Ether channels from the access layer.

The attached drawing shows the topology's I was thinking of however, I have some concerns which I've documented in the drawing. Any advise would be appreciated. 

Andy

[Richard Bradfield]

Andy,

I would prefer your first design, with the Dist switches in different buildings, why? because if you have a stack just in one building, then most probably you  would be running the fibre to the Netgear across the same 6 core fibre cable, so if that got damaged you would loose connection to that particular stack. Also I note you do not have any direct connection between the Netgear switches if you did then there is a redundant path to each Netgear switch. The only thing I don't know how spanning tree protocol works on the Netgear switches.

[networkwise]

Thanks for the input, very helpful.

Appreciate it..

Andy

[Joseph W. Doherty · Accepted solution]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

Would you consider using a chassis, with redundant sups, power supplies and downlinks, to same device, on different line cards, as a distribution (or core) device?  If so, that's pretty much what you have with a stack (or VSS pair).  I.e. if a chassis (or VSS pair) is a valid design, why wouldn't a stack be too.

The advantages you noted, on your diagram, for a stack would also apply to a chassis (or VSS pair).  Plus, you often get more bandwidth.  STP doesn't need to block a redundant L2 path and stacking cables may provide more bandwidth then "normal" port.

Etherchannel should work between vendors using LACP.  (At least on Cisco, don't forget to chose an optimal hashing algorithm.)

The main disadvantage of a Cisco stack, like a single chassis supporting redundancy, all the hardware is in one physical location.  Something happens at that single physical location, you lose that distribution node.

Another disadvantage, you're more subject to a software fault as there's one logical device (also applies to single chassis or VSS pair), but with multiple L2 devices, you're then subject to faults in FHRP, STP, etc.

[Jon Marshall]

-

[visitor68 · Accepted solution]

Andy, you've been given some good advice, I'll just add a comment or 2.

I am no fan of stacking because it creates a single point of failure. If the stack experiences a glitch, your whole stack is shot and all the access layer switches are dead in the water. The stack mitigates your redundant design. While it's true that stacking will allow you to create MLAGs (EtherChannels) from the access layer stacks and allow you to have full bi-sectional bandwidth, with such low-end gear in place and OM1 fiber, something tells me that their traffic requirements are not so great. I think a blocking architecture may be sufficient. At least you will have unmitigated redundancy.

Now what I am about to say is going to be blasphemous on this site, but there are vendors out there who offer 10G switches that offer a VPC-like technology for the cost (or close to it) of a 1G Cisco 3850. for example, I deployed Dell Force10 s4820 10G switches with 10GBaseT interfaces and created a VLT domain (like Cisco's vPC), and I did it for about $12,000 a switch. It had 48 x 10G ports + 4 x 40G ports. I believe Dell also has their N-Series 1G switches that also offer VLT, which is normally a 10G technology. So, don't pigeon-hole yourself.

My 2 cents...

[networkwise]

Thank you Joseph,

Appreciate your input. All the input to this question from you all has been excellent. Very helpful in guiding me to a decision.

Thanks

Andy