Home > IT Answers > Networking > Connecting data closet switches to the core w...
NetworkingATE
1,545 pts.
 Connecting data closet switches to the core with redundant links
Networking, Routing and switching, Switches, Core, core networking, Subnetting, VLAN, spanning tree, Cisco switches, Cisco 6509, VSS, OSPF, Protocols
I'm moving our flat network which covers a three-story building and about 450 users to a subnetted network. I would like to know about connecting data closet switches to the core with redundant links. Can I implement OSPF on the core and connecting switches and eliminate spanning tree? In this environment do I really need VLANs per subnet or can I just subnet? I'm starting with new Cisco 6509 switches running VSS. I'm using the new switch to route between subnets. All data closet switches and server switches will have redundant links back to the core.
ASKED: Jun 2, 2009  1:15 PM GMT
UPDATED: June 2, 2009  11:33:44 PM GMT
RELATED QUESTIONS
CPartsenidis
180 pts.
Answer Wiki:
My understanding is that you are looking into a full-scale redesign of your network.

The most important factor in these cases (and especially when we are dealing with network as large as yours) is the network backbone and available equipment.

The Cisco 6500 series Catalyst switches provide a great amount of flexibility for your design, and I highly recommend you make use of its available features.

Firstly, you can surely survive without OSPF. Since your network backbone will consist of a number of switches connecting back to your core switch (6509), there is no need for OSPF. Static routing will do the job just fine, and once you have the network up and running without problems, you can consider adding more complexity (e.g OSPF) if you still believe its required.

Without having a full view of your network, available equipment and requirements, here's what I suggest:

Each floor has a number of switches for workstation connectivity. These switches will connect to a central floor switch e.g 3560 or 3750 that should have at least two fiber optic links to your core switch, for redundancy. If you have switches available, you can even wire them directly to your core switch.

Without second thought, VLANs are a must for your network size - that's one subnet per VLAN. Don't think about it twice! If you include Per-VLAN Spanning Tree, you'll be able to have the full benefits of Spanning Tree and VLANning.

In addition, it would be wise to also setup services such as VTP and VTP Pruning, to ensure your VLAN data is replicated correctly amongst all member switches and that trunks carry only the necessary data, depending on their location and configuration.

If you would like to read more on VLANs, Spanning Tree Protocol and how you should segment your network, you can browse through Searchnetworking.com's ATE section and visit www.Firewall.cx.

Hope that helps!

Chris Partsenidis
Last Wiki Answer Submitted:  Jun 2, 2009  3:27 PM (GMT)  by  CPartsenidis   180 pts.
Latest Answer Wiki Contributors:  BlankReg   11,280 pts.
To see other answers submitted to the Answer Wiki View Answer History.
RSS - Discussions Help
Discuss This Question:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _



 

I did post an answer to this, but Chris seems to have posted at about the same time, and has overwritten what I put.

I basically said the same as him, but wondered if youy need any OSPF or static routes at all, as I would think that all the routing can happen on the 6509s, and not at the edge. Almost all designs are still with centralised resources, so routing at the edge is an expensive option. Go for L2 switches at the edge, with VLANS, and route at the core. As all the subnets are on these switches, no additional routing protocols are needed.

Also in this scenario, the servers should connect to the 6509, and not to another switch. Otherwise they all share a much smaller bandwidth to the rest of the network.

My other advice would be to use the redundant uplinks as aggregated links, and gain extra bandwidth, at no extra cost (other than a bit of config). Then, with a little bit of careful design, you don’t need to worry about STP, and all you lose is speed if a link fails. Double the pleasure and almost no pain !

I presume the redundant links use diverse routing to the core ? Otherwise there is definately no real benefit in using them as true redundant links, as the failures are most likely going to be due to a physical problem, and it is very likely to effect both if they share the same physical route.

BlankReg
 11,280 pts.