As far as I know the Deduplication Services and the Deduplication Storage Folder cannot be configured in a cluster aware manner and yes you are correct you should be using Optimized Duplication between two media servers instead. A traditional cluster still gives a single point of failure (the disk array), optimized duplication results in two separate storage locations as well as two separate media servers, reducing the single point of failure
With regards clustering BE without Dedup, whilst we still provide the function, normally a CASO setup might be better as it can be used to load balance and provide resilience for continuing to run backups (as in a CASO environment different media servers can be used to backup the same data if the usual server has failed) If you add deduplication and optimized duplication to the CASO setup you also get the resilience for the restore.
In most environments you would try to plan the setup so that the likelihood of loss of production server at the same time as the backup servers is miminized in such a way that you should have time to recover a missing backup server before it becomes issue and whilst we do still see clustered BE servers it has become one of the lesser used configurations over the last few years.
BTW clustering the CAS server itself if you have a large enough environment to require using a FULL SQL instance for the CAS BEDB is not really possible, which is potentially a limitation, as currently if the CAS fails then you may need to know how to revert the MBE servers to stand alone while you fix the CAS.
