From my experiance with our arrays, this seems like a bad idea to create one big lun. You want to a few luns so that the processor boards in the storage array can be used to their fullest extend(also qdepth will be an issue). If you have one lun, it will only be assigned to a single processor board. Rather create a few luns, typically I dont create luns bigger than 2.5TB. Rather create a couple of 2.5TB(as a max or lower) luns and use something like Datastore Clusters in vmware to spread the workload.
If you'd like to see the optimal number of luns/paths to use you can configure your array in CPK, which can provide you some guidance on how to accomplish it.
You can use dynamic tiering, based on the policy you use, it will move the data based on the cycle times specified.