ATP

[With Liviu Iftode, Associate Professor, Dept of Computer Science, Rutgers University, USA]

For more details, visit ATP Project Web Site

In a true ubiquitous environment, users should be allowed to change networks and hosts seamlessly and communication should continue even if the user is not available for a short period of time. Current Internet architecture offers a limited support for ubiquitous communication. In traditional TCP/IP, the connection is identified by the IP address of the end hosts and the user is bound to the same host during the connection lifetime. Although the mobile IP protocol provides a solution to the host mobility between different networks, a user is bound to a single host during the lifetime of a connection. As ubiquitous computing emerges, the user should become the focus of communication and not the end hosts.

 To achieve this goal, connections should be carried between users, independent from the host on which the user is located. Peer-to-peer lookup services provide a mechanism to map a content into a specific host and to query the location of this content in a peer-to-peer (P2P) network. Considering a user's endpoint as a content, a transport layer protocol over a P2P network uses the lookup service mechanism to locate the end hosts. The challenge is how to maintain a reliable connection between the users' endpoints as they roam in the environment moving from one host to another leading to a dynamic change of the user ID (content) to host mapping. In such an environment, data for the same connection can be produced at different hosts and consumed at multiple locations depending on the location of the user endpoint. In TCP , this mapping is static during the lifetime of the connection.

In this project, we propose the Autonomous Transport Protocol (ATP). In ATP, autonomy allows dynamic endpoint relocation on different end hosts without disrupting the transport connection between them. The ATP has the following features:

 It does not enforce any naming scheme on the user application. The application is responsible for uniquely identifying the endpoint.

 The endpoints of a transport connection are defined as contents in the P2P network. This decouples the connection from the physical host where the user endpoint is located, and hence ensures autonomy.

 Mobility of the endpoints is handled via the P2P network by dynamically changing the mapping between the endpoint and the host. The ATP layer is responsible for moving segments to the destination and the acknowledgment to the source regardless of their current mapping in the P2P network.

 Since a P2P network is built as an overlay network, the ATP layer in the intermediate nodes between the source and destination endpoints can actively participate in the connection, for example, to buffer data for the destination endpoints during periods of unavailability.

 Data is transferred by a combination of active and passive operations, where the ATP layer of a node can decide whether to actively push the data to the destination or to passively wait for the destination endpoint to pull the data. The decision to whether to use the active or passive modes can be taken by a local policy on the node running the ATP protocol.

Current research directions include designing and evaluating different policies for the pull/push decision, designing and implementing applications over ATP, and enhancing the protocols to increase system security.