Cluster access control
Once you have the shiny new cluster up and running, you need to consider questions of access and security. Who can access the data on the cluster—is there sensitive data that you really don't want the whole user base to see?
The Hadoop security model
Until very recently, Hadoop had a security model that could, at best, be described as "marking only". It associated an owner and group with each file but, as we'll see, did very little validation of a given client connection. Strong security would manage not only the markings given to a file but also the identities of all connecting users.
