W. Curtis Preston recently authored an article on Searchstorage.com.au explaining global deduplication. This is an important topic which frequently causes confusion. Curtis does a good job explaining the technology and what it means to end users and I recommend the article.
A quick summary is that global deduplication means that a common deduplication repository is shared by multiple nodes in a system. In these environments, a customer can backup their data to any node on a system and it will be deduplicated against related data. This provides improved ease of use and scalability.