IBN Layer

We propose a new Instance-Based Network (IBN). The proposed network provides the flexibility of having different instances of the same content. It is up to the user of the IBN network to define the relation between these instances. For example, a file archiving system can use the new IBN to store different versions of the same file. Here, the file name represents the common content and the different versions represent the different instances of the same file. A user of this file archiving system may want to query for a particular version of the file (instance) or ask for the latest version of the file. This feature is unique in the proposed IBN compared to any P2P lookup service.

Features/Goals

The proposed IBN (whose components are shown in the above figure) extends the functionality of the current peer-to-peer lookup services, which provides a mechanism to map a content identifier to a node assigned by the IBN. These extended functionalities are:

	Content-node mapping: The IBN user can ask the IBN to map a content to a particular node (through a publish operation). All mapping are leased, that is, if the user does not refresh the lease before it expires, the content is removed from the IBN.
	Content communication: Application endpoints, defined by contents, can send messages to other content-identified endpoints.
	Instance-based routing: The IBN can route a message to a specific content instance or to the nearest instance (the neighborhood metric is defined below) if no exact match is found for the destination content instance.
	Replication: The IBN replicates the stored contents in order to provide fault tolerance.
	Caching: Nodes along the query path can cache a content to provide fast answers to future queries.

Addressing

A content of the new IBN is addressed using a name X and an instance identifier (i1, i2, …., in), where i1, …, in are n integer numbers (as shown in the above figure). We use the notation (X : i1, …, in) to refer to an instance of a content X. The semantics and dimensionality (n) of the instant identifier tuple is assigned by the user of the IBN network. These semantics include the ordering relation between different instances. For example, in a file archiving system, a file name can be represented as (logfile : 1; 0; 1) to represent the version 1.01 of the file logfile. These semantics are assigned by the file archiving system. The user of the IBN network maps a content instance to a particular node. The next section describes how the routing is performed in the proposed IBN.

Routing

The routing in the proposed IBN network is instance-based. A message destined to content (X : i1, …, in) is routed to the nearest published instance of the same content X to the destination instance. The routing algorithm of the IBN is illustrated in Algorithm 1. The function Closest in the algorithm represents one possible ordering relation between different relations. This is the comparison function used by the ATP protocol. The proposed IBN extends the routing techniques of the current P2P lookup services. Appendix B shows the API that the IBN network exports to upper layers.

[ IBN Layer ] [ ATP Layer ]

Department of Computer Science

Last changed: February 20, 2004

Web Accessibility