next up previous
Next: Virtual Network Configuration for Up: Network Control Protocol Overview Previous: Physical Layer of the

ATM Network Configuration Layer

This section briefly describes how ATM VCs are setup by the NCP. As the orderwire network determines the topology of all nodes in the wireless segment (e.g., RNs, ESs) in our architecture, and establishes link connectivity among adjacent nodes, setup is still required of the actual ATM circuits on which wireless ATM are carried on the user data overlay network. This is accomplished by providing standard ATM signaling capabilities to RNs and ESs and using Classical IP over ATM [13] to associate ATM VCs to IP addresses. The Classical IP over ATM implementation provided works for PVCs and SVCs (using ATMARP). Since an ES may connect to multiple RNs (wireless connections) or ATM switches (wired connections), it can be thought of as a software-based ATM switch. In this sense, an ES features ATM PNNI signaling while an RN features ATM UNI signaling. By default, an RN creates one wireless-ATM protocol stack and establishes an ATM VC signaling channel on such a stack; however, the stack is initially in an inactive state (i.e., non-operational mode) since there is no link connectivity to another node established yet. Likewise, an ES creates a predefined number of wireless-ATM protocol stacks - acting like ports in an ATM switch - and establishes ATM VC signaling channels on all configured stacks which are also initialized as inactive. Wireless-ATM protocol stacks are controlled by a daemon, called the adaptation manager, which acts on behalf of the orderwire network. The adaptation manager daemon not only controls the stacks by setting their state to either active or inactive (default), but also may modify configuration parameters of the stacks to provide dynamic adaptation to link conditions. Two possible scenarios illustrate the interactions between the orderwire network and the wireless-ATM network. In the first scenario the orderwire detects link connectivity between an adjacent pair of nodes (e.g., RN-ES or ES-ES). In this case, the orderwire network requests an inactive stack from the adaptation manager daemon at each end and associates them with a designated address. Upon establishment of link connectivity, a requested wireless stack has its state set to active and is ready to operate. Note that since the signaling channels are preconfigured on the stacks in question, users on the wireless establish end-to-end connections exactly as if they were connected in a wired ATM network. The other scenario occurs when the orderwire network detects a broken connection, at the link level, between two connected nodes. This case is typical of an RN moving away from the connectivity range of an ES. The orderwire network thus contacts the adaptation manager daemon at each end to set the wireless stacks in question to inactive. Since a wireless stack is never destroyed, it can be reused in a future request from the orderwire to establish connectivity to another pair of nodes.


next up previous
Next: Virtual Network Configuration for Up: Network Control Protocol Overview Previous: Physical Layer of the

Steve Bush
Tue Nov 12 11:48:31 CST 1996