Always On is a commercial name for a set of High Availability features that include Failover Cluster Instances and Availability Groups.

The main difference is that a clustered instance has the same binaries installed and configured on two or mode cluster nodes (physical or virtual machines) and the database files are sitting on a shared disk. Only one node at a time can own the data disks and run the SQL Server instance (active node) and all other nodes are said to be passive. When the node fails for any reason, the disks are moved to one of the passive nodes and the service is started, making it the active node. All clients connecting to the clustered instance use its virtual network name, which is moved to the active node, ensuring that all clients are able to connect to the active node.

With Availability Groups, two or more copies of the same database are synchronized across multiple nodes. In this case you don't have a single copy of the database files, but you have multiple copies (replicas). One of those copies is read/write, while all other copies are read-only. To ensure that the clients are connecting to the appropriate replica, a virtual netkwork name (listener) is used, which is similar to what the cluster virtual network name does. With an additional connection string property, you can control whether you're hitting the read/write or read-only replica.

For more information, please refer to the stairway to HA on this site starting at this link[/url]