The complexity of crash failures

A crashing network protocol is an asynchronous protocol whose memory does not survive crashes. In [. TVW],it is shown that a crahing network protocol that works over unreliable links can be driven to arbkrary global states. [n this paper, we further characterize the reachable states caused by crash failures using other link models: reliable non-FIFO and reliable FIFO links. We show that with reliable non-FIFO links any acyclic subset oj nodes and links can be driven to arbitrary states. We show that with reliable FIFO links, only nodes can be driven to arbitrary states. Taken together with the earlier result, we show a strict hierarchy in terms of the reachable states caused by crash failures in the three link models.