5500 to 4500 migration

Unanswered Question

Attached is our existing network consisting of 5500. The plan is to replace them with 4506 chassis.

Referring to the diagram

Block 2 switch is acting as the VTP server and is connected to the edge routers

All other 5500's trunk to the Block 2 switch and hence VLANs are spread across

SVIs are created at each 5500 and are not centrally created on the block 2 switch

To connect at IP layer, all the RSM switches run EIGRP to learn the subnets

VLANs need to be spread across all the blocks even after this migration as there is dependency.

Should i be creating SVI's at each switch and then EIGRP again to connect them at Layer 3 or i would be better of creating all the SVI's in the block 2 switch and just have layer 2 trunks from other 4506?

What are the Pro's and Cons of both the approach?

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Jon Marshall Sun, 09/09/2007 - 22:45



It's not clear from your diagram where the switches are physically located. If two of the switches could be located within the same LAN room i would look to use 2 of the new 4500 switches to create a core and then attach the other switches with layer 2 trunks.

The SVI's for the vlans would be created only on the 2 4500 core switches and you could run HSRP between them for redundancy. You would need to to have a layer 2 trunk between the 2 4500's.

Each of the other 4500 switches should then have a layer 2 connection to each of the 4500 core switches for redundancy.



All the 5500 switches are in different blocks but in the same campus.

The 5500 in the middle (termed as blck 2) is the core and hence when i will be replacing them with the 4500, i will be having a single core.

My plan was similar to have the 4506 at block2 as core and create all SVI's in it and just have L2 trunks to other 4506.

B ut was just wondering why the RSM's were configured with SVI's at all the blocks and a L3 link was created between them.

Was this a goo approach or was it only done to distribute the load in some way?


Jon Marshall Mon, 09/10/2007 - 00:05


It would have made more sense if the uplinks between the block 2 switch and the other switches were layer 3 links. If that was the case then it would in effect have been a routed design.

I'm not that familiar with the 5500 switch so maybe layer 3 switching was more intensive on the CPU and memory but if you need your vlans to extend across all switches i would recommend having the layer 3 SVI's on just your core switch.


glen.grant Mon, 09/10/2007 - 04:33

I see your question also why they had rsm's in all switches if one was the vtp server doing the routing . Were the same vlans across all the switches or were they isolated to each individual switch , routed between rsm's.There was no added overhead using a RSM card specially if you had configured mls switching , we ran our whole distrubution layer for a big corp on these and usually the cpu was less than 10% except in a couple of cases where there were viruses like slammer where it bombarded the cpu with unknown addresses. There is no reason why you can't have the 4500 as the core but if you do this I would have redundant supervisors in it . Use that as vtp server and use the other switches at layer 2 . There are certainly numerous designs you could make work but make sure you some type of redundancy built in .


The VLANs are spread across all the switches but the SVI is not created only on one switch but they are also distributed across.

Now to route these interface vlans, EIGRP was run between them. i do not know whether this method was deployed to reduce the overhead on one RSM.

It doesn't seem to be a good design to have both a L2 and a L3 link between the switches.

When we replace with the 4500, there would be 2 of them in HSRP thus providing redundancy


Joseph W. Doherty Mon, 09/10/2007 - 06:59

This isn't directed to your initial questions, but was wondering whether you've reviewed the rest of Cisco's product line? Gig Ethernet is starting to become popular, more so I would expect over the next couple of years. One possible concern with the 4500 series is the 6 Gbps per line card.


This Discussion