You can have iterations running in parallel that are working on features that are not tightly interrelated. As long as the two iterations are not overly dependent upon each other, there should be no issue running them in parallel. Practically speaking, you would want to share roles like solution architect, DBA, etc. across iterations, as all iterations are all working with the same architecture, same DB, etc. but don't need full-time positions for some of these roles.

The recommendation is NOT to slightly overlap iterations. The reason is because you lose the core value you receive from iterative development: the feedback from one end-of-iteration review should be used as input into the planning for the next iteration. If the iterations overlap, you lose the chance to incorporate the lessons learned, customer change requests, etc. and also any new priorities appearing from the customer reviews. Otherwise, your first chance of incorporating the feedback from iteration #1 would be iteration #3. This also means that you would lose the chance to incorporate feedback from the last two iterations unless the project goes over schedule.

So, if you are going to do overlaps for resource balancing purposes, overlap in parallel.