Putting together the network as you describe it with no single points of failure is very much possible, although not easy. There are a wide range of single points of failure you will need to deal with, from physical diversity of your two communications links to picking which boxes plug into which UPS. Note that unless your users are also dual homed to your infrastructure, the LAN link to each box will remain a single point of failure as well as the end user boxes. Your servers and the data on them could also be a single point of failure, if they are not implemented as cluster systems with redundant LAN links to both of your switches.
Load balancing is also possible, but be aware that when it comes to load balancing, the general rule is "load balancing -- robustness -- cost/complexity -- you only get to choose two out of three."
DECnet was handling redundant setups effectively while TCP/IP was still emerging. The techniques you will need to support redundant DECnet are similar to the approach you will need to support high availability TCP/IP, just the names of the protocols are changed.
As for some rough info on how to do it, you can't even scratch the surface in a brief e-mail. I wrote a book about it "High Availability Networking with Cisco" and barely scratched the surface. Grab a copy (used ones are selling cheap) and spend some time doing some heavy reading... You'll find that while the text may put you to sleep, the concepts and concerns raised will keep you awake nights worrying about all the aspects you haven't covered yet. If this is a serious application, you have a lot to learn in a short time.
Good luck and have fun!
Vincent C Jones
www.networkingunlimited.com
