RSTP Scaling and redesign

Unanswered Question
May 2nd, 2007

Hi,

Im looking to redesign the current network and have come across quite a few issues with the current topology.

Currently the network looks similar to this.

|-2950-|

|-2950--2950-|-2950-|-2950

|-2950|

7206-6500-2950-|-2950|-2950-|

| |-2950- 2950

7206-6500-3500

|

Colo

The 7206s connect to the upstream ISPs and also run OSPF to the 6500s. The real issue is with the design of the 2950s. In a LAN environment there would be more options but the 2950s are all connected via RF P2P Ethernet and are anything

between 2-5km apart. The actual no of 2950s actually is a lot larger than the above diagram. Customers connect

to each switch and are VLAN'd back into the core and the number is growing.

Issues are:

* RSTP 802.1w is run in the 2950 network. There are obvioulsy limitations with increasing csr no's/VLANs and scalability.

* The possibility of broadcast storms with so many possible STP rings is keeping me awake!

* Each site is allocated a /24 for customers.

The design above is neither scalable in terms of design or IP allocation.

The above needs to be redesigned and the 2950s need to be replaced. EoMPLS (or ATOM) has been discussed as

a possible saviour. In my opinion the benefits of this will be negligible in the current design due to the centralised core and the depth of

the L2 network. I may be wrong. MSTP is also an option but migration will be difficult and also doesnt address the added requirement of QoS and other possible value added services (such as Q in Q).

Any ideas to possible redesign and hardware upgrade would be welcome.

Thanks.

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.

Actions

This Discussion