Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

We are constructing a new networking infrastructure to replace out 1GbE backbone and have decided upon using 4x Dell PowerConnect 8024F's as our "core" switches.

As per the diagram below we have 2x 8024's upstairs and 2x downstairs providing links over MMF fiber for redundancy. Data being transferred is a mixture (70/30) of iSCSI/LAN on separate VLANs.

Network Diagram

How can we best configure these switches to allow for redundancy and throughput, 2x stacks of 2 8024's or 1x stack of 4 switches and where should LAG be employed?

share|improve this question

2 Answers 2

up vote 1 down vote accepted

I've been bitten by weird failures of both Cisco's and Dell's stacking technologies that I will not depend on them to provide redundancy.

The craziest example would be a stack of 3 Cisco 3750's where two switches appeared to have failed. After a protracted support call with Cisco, we figured out that the problem was the third, seemingly functional, switch. If I removed the third switch from the stack, the other two switches came back to life. And no, the third switch was not the master. But simply having it connected in the stack ring caused the other switches to stop working.

With a couple of experiences like that, I now design network topologies with these points in mind:

  • Assume that an entire stack fails instead of just one switch in the stack. Do not depend on the stack ring for redundancy.
  • Connect all the stack members with at least one Ethernet cable to another switch not in that stack. That way if the stack ring fails or members become separated, spanning tree takes over and you still have access to the devices connected to the switch via this backup link.
  • If you LAG the backup link, do not split this LAG across stack members.
share|improve this answer
    
So basically, for our situation - avoid stacking and instead have identical configurations on the 2x related switches and LAG the connections between all core switches? –  Myles Gray Jun 21 '13 at 14:56
    
That would be my suggestion, yes. –  longneck Jun 21 '13 at 16:09

Stacking configuration for the upstairs and downstairs pairs of 8024's gives you a huge advantage in ISL link bandwidth without having to sacrifice any of the 24 ports available on each switch for a LAG connection.

Stacking also provides a bit more ease of management, but that's beside the point w/ your question.

The problems mentioned by longneck are definitely one-off scenarios, and are not something you're likely to run into with these newer switches (especially if you stay somewhat up-to-date on firmware). In comparison to a LAG link though, a stacking link probably isn't going to provide the same fault tolerance (single connection vs multiple links).

If redundancy is a big concern and you have the links to spare, I would agree with longneck on suggesting the LAG connections.

share|improve this answer
1  
Agreed. If you had 4 pairs of switches instead of 4 switches, I would be comfortable with having 4 stacks. (Assuming, of course, that you need the redundancy of 2 aggregation switches both upstairs and downstairs.) –  longneck Jun 21 '13 at 18:42
    
The 8024's don't have dedicated stacking IFs or expandable backplanes, you must sacrifice some of the 24 ports to provide stacking or LAGing, on that premise then it seems LAGing is more advantagous? –  Myles Gray Jun 23 '13 at 12:53
1  
My mistake - was thinking this model had QSFP modules for stacking like the 81xx series. Sounds like LAG connections will in fact be best in your scenario after all. –  JimNim Jun 24 '13 at 14:10

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.