Be advised that though the policy sharing and the VIP failover aspects of clustering sound great on paper, you do currently lose some detailed logging with this.
I'm clustered and we had some major freaking issues in deployment where a hardware issue caused the cluster to flap all over heck.
I'm friends with some folks who've recently done a much larger deployment and they are not clustering. And their deployment went much more smoothly for whatever it's worth
I believe the clustering setup is covered ... poorly.. in the product guide. We had (quickstart) consulting, SE during proof of concept, and a lot of support help with it as it surely wan't documented well enough to forge ahead on ones own, if memory serves.