redundancy and load balancing between catalyst 3524-XL and catalyst 3548-XL

Do you have to trunk? Why not turn up a diff subnet on the remote side and turn up layer 3 routing between sites via both links? If one drops, routing will failover to the second link since the adjacency will drop. If you can't do this, I'm not sure why a link failure wouldn't failover with your configuration anyway. If you have two trunks, only one is being used at any given time (to avoid a loop), and if it fails, spanning-tree should run and it should go forwarding on the other trunk. What am I missing?

If you want to load share, and presuming both links are equal bandwidth and go between the same two switches, have you tried etherchannel? If its just a layer 2 connection at each switch, they most likely have no idea that there is a microwave between them, and etherchannel would allow both ports to belong in the same channel unless one went down. Else the layer 3 suggestion from my previous post will work fine, as long as both links are equal bandwidth, any decent routing protocol should load share the paths presuming its the same end points on both sides.

routing is not possible, for "historic" reasons i have to stick to the current setup. the microwave links behaving transparent to the switches and the switches "believe" there is a cable between them. yes both links have the same capabilities, 34mbit/s. i had tried just to enable all for interfaces and leave spanning tree do the job. i noticed it takes about 5 minutes for traffic to pass if the currently active link fails. so i was looking for something which will fail over within seconds. i believe port group is etherchannel on cisco switches, if not please elaborate, i try in the meanwhile to find some documentation on the web.

thanks to both of you.

Jan

I think that the key here is 'how realistic a wire does the microwave link provide'.

If it really does look like a wire to the switch and the port goes DOWN iif the link does then I think that you could use etherchannel. This uses port groups in its confguration statements.

If on the other hand it is not such a good simulation of a wire you will I think have to use Spanning tree. You can do a form of load balancing with STP if you have more than one VLAN in use by having some VLANS use one link by preference and some VLANs using the other.

By defaut STP takes 30 (35?) seconds I seem to recall to transition to forwarding in the event of a link failure in your topology. Thre is now RSTP but maybe your old switched do not do it.

The timers are tunable to make convergence faster. Thing is you say that yours is taking 5 mins.

Unless your timers have been changed the other way then something else is causing the delay.

If you post the whole config maybe someone will have a look?