We have a 48 port 2960G switch distributing our LAN to numerous 24 port 2960G switches.
We have been talking about bringing in another 48 port 2960G switch for redundancy.
Can someone please let me know what the best design options are for this setup, in terms of redundancy and possibly even load balancing? I am interested to find out what Cisco technologies/features are recommended in this situation.
I created a quick drawing of what we currently have and what we are thinking about. None of these devices have been put into production yet. Let me know if you have any questions, thanks.
I recommend running 4 links from each access switch, 2 links going into each core switch.
Configure the dual links going into each core switch with etherchanneling.
Etherchanneling provides LB/FT.
Read all about Etherchanneling @
Additionally, configure each core switch as spanning-tree vlan # root primary | secondary.
Please rate helpful posts.
Thanks Edison, but a few questions...
Why no link to connect the two core switches together?
These are all gigabit connections, so I think the "dual links" might be a bit overkill in our specific situation. Besides, I might have a few access switches that need to have all the ports populated already as it is.
Would the UplinkFast/BackboneFast features be something that I should look into as well?
Yes, I forgot interconnecting the 2 cores.
Go with 2 links with etherchanneling configured.
As for uplink/backbonefast, I recommend enabling uplink on the access switches and backbonefast on the root switches (cores).
Have a read here:
Please rate helpful posts.
First and foremost, good thinking :-) Its good to have redundancy.
1. Two uplinks to the distribution switches from the access layer switches.
2. STP - Hardcode your bridge ID and make sure that for certain vlans Distribution Switch A is the primary and for certain vlans Distribution Switch B is the primary.
3. One port from the access layer switch is going to be in forwarding mode while the other one is in blocking state. In case of the primary path failure, Spanning Tree will recalculate the path to the root bridge and start forwarding traffic.
4. Also choose the right STP Technology for your environment.
Please refer the below docs for some good information on designing campus networks.
Rapid Spanning Tree Protocol:
High Availability Campus Recovery Analysis
Ok, scratch all that. My project has been updated, to the point where my design needs to go back to the drawing board. I was told to implement 100% end-to-end redundancy (from the distribution switches to the servers).
We have nine racks of servers for this project. Each rack will get a primary access switch, and a secondary access switch. That is a total of 18 access switches. These switches will all have uplinks to two distribution switches. For the servers themselves, we will utilize NIC adapter teaming. The server network adapters will be configured in a primary/secondary relationship. I have an example drawing of how I think one rack would be laid out.
- The two distribution switches are 2960G, 48 ports (already purchased).
- The access switches are 2960G, 24 ports (already purchased).
- There will be a total of five vlans, but only two of them will carry around 90% of the traffic.
- The network will be strictly data. There is no voice traffic.
- There will be a ballpark of around 60 servers (no more than 10 servers on a rack). Almost half of these servers are only there as backups.
- There will be no end-user workstations off any of these switches. This is strictly a server farm.
- I do not have a strong feel of the overall traffic on this network, but I was told that it could pass a few terabytes per day (moderate-low load I suppose?).
I have a number of questions....
1. Can two 2960G switches effeciently handle the stress of connecting to 18 access switches each? There will also be other links on these distribution switches, such as connections to the internet and end-user access from our corporate network.
2. All five vlans will be propagated to all of our racks. I was hoping to balance the traffic by vlan. Make one distribution/access switch the primary link for one vlan - while making the other distribution/access switch the primary link for a different vlan. Is that feasible under my circumstances?
3. Etherchanneling the trunks (that is, having two or more links from a distribution switch to an access switch) would not be possible, because you can only configure six etherchannels per switch total. My distribution switches would have 18 different etherchannels (one channel for each access switch). Am I right in thinking this?
4. Will NIC adaptor teaming on each server effectively work in this scenario, or is there a better solution for redundancy? I read this can be done with Cisco's Link-State Tracking feature.
5. Are there any modification suggestions to my drawing, or is that the best end-to-end redundant method for my situation?
Feel free to ask questions. I look forward to some feedback, and appreciate the help.
What you have is exactly what we want to accomplish. Right now we have multiple swithes in about 30 closets and they are all connected to a CAT6509 Ver 7.1. We recently purchase another 6509 with sup 720 12.2(14)SX. Now I wanted to run two fibers to both 6509 from different closets although it was not tested yet, I was told by a cisco engineer that every time a link or a power outage on a closet, the whole network will go down to recalculate STP which would take about 50sec. I was told that using Rapid Spanning tree or layer 3 switching is the best solution. Our problem is that most of our switches in the closets are 3500XL and doesn't support Rapid Spanning Tree. My other thing was we just got 3 3750 all stack and wanted to run them both in our 6509 using Rapid Spanning Tree. With the new 6509 with SUP720 and the 3750 the process looks simple just enabling Spanning tree mode to rapid pvst will do the work but with old 6509 I cannot see that command. Do you also think that If I combine both pvst+ and rapid pvst will work the same? Let say a link from the 3750 goes down will the PVST+ recalculate STP even if the segment that went down is using Rapid-PVST? It would be nice to have all those equipment just for testing, unfortunately we can't do anything in live productions.
i'm attaching a ppt in which i have shown only one rack design .Distribute the servers according to ur requirement and give all the links between all switches as trunk port so that they can allow all the vlans.here dashed lines are secondary link and dark lines are primary link .pls get a view and if any improvement then do that and revert me back.i suppose this will fulfill ur requirment.
one thing more i think u should use 4500 or 3750 series for distribution switch.