Error control in multi-layered stacks

The decomposition of networking functions into smaller subsets simplifies implementation and leads to open standards. Nonetheless, in applications such as mobile wireless ATM, where resources are scarce, protocol stacks must not compromise on performance. The layers of a protocol stack progressively enhance and add reliability to the supported services. Along with additional new functions, the higher layers frequently implement error control algorithms to overcome residual errors that escape the lower layers. If relatively large numbers of residual errors escape, as is likely in the mobile wireless ATM environment, then frequent retransmissions or reinitialization of the higher layer protocol entity could result in drastic loss of throughput. In this paper we examine the consequences of errors spanning two layers of a protocol stack using an explicit model to account for the interaction between two layers.