Due to the very nature of peer-to-peer, it cannot be assumed that nodes within such a network are connected at all times. For example, a computer connected to the Napster network might be switched off when it is not in use. Consequently the MP3’s that it was making public are no longer available on the network (unless provided by another node). When the computer is switched back on and the network connection re-established, the MP3’s once again become available. Similarly with a computational based application it cannot be assumed that a node can always work on a computation, and thus this needs to be taken into account for distributed computations.
As with the connection bandwidth, the intermittent connectivity of a node can play an important role in what network architecture is used and can have a significant effect on a systems perceived dependability. When designing a peer-to-peer system, it is important to cater for a node’s intermittent connectivity and not assume that the node will always be connected.