An Innovative Approach to Content Search Across P2P Inter-Networks Potharaju S.R.P Saradhi Mohmed Nazuruddin Shaik Potharaju S R Aditya Under The Guidance of Mr. P. Barath Asst.Prof, SCS Vellore Institute of Technology

Definition A distributed architecture consisting of a collection of resources performing a distributed function is called a peer to peer architecture. P2P computing is the sharing of computer resources and services by direct exchange between systems.

Aim To effectively use peers to share their content with topologically closer peers. To reduce the number of redundant messages in the network without compromising on the search performance. To effectively utilize the topology of the underlying network to increase the performance of the query routing algorithm.

C A B Problem with the content search strategies in the network Packet

Approach To design a new algorithm for converting any physical network into a conceptual network. To implement a new query routing algorithm in the derived conceptual network to locate data present in the actual physical network.

Proposed System Architecture Peers in the system act as client peers and super-peers in different hierarchies. Super-peers act as local search hubs, proxying search requests on behalf of these peers. Super-peers are connected with each other and organized amongst themselves into a backbone overlay network on the super-peer tier.

Multi-tier Architecture

Advantages of Proposed Architecture The hierarchical structure of this system combines advantages of both centralized and pure P2P systems. It combines the efficiency of a centralized search with the Autonomy Load balancing and Robustness provided by the distributed search mechanisms.

Advantages of Proposed Architecture (Contd...) The proposed architecture distributes the load on the central server to many super-peers. No single super-peer is required to handle a very large load. Nor will one peer become a bottleneck or a point of failure for the entire system.

Selection of Nearest Super-peer Let dist(p,q) represent the underlying distance between peers p and q. Let dist RTT (p,q) to represent the detected RTT value and dist TTL (p,q) to represent the detected TTL value. Let d be a predefined value to denote the latency difference of two paths.

Selection of Nearest Super-peer (Contd...) Given peers p, q, s, t then the distance can be compared using the following concept: If |dist RTT (p,q)- dist RTT (s,t)|>d then If dist RTT (p,q)≥ dist RTT (s,t) then dist(p,q)> dist(s,t) Else dist(p,q)< dist(s,t) Else If dist TTL (p,q)≥ dist TTL (s,t) then dist(p,q)> dist(s,t) Else dist(p,q)< dist(s,t)

The Proposed Neighbour Information Based Algorithm * A i : The identifier of peer i. *hop t : the network distance that is t hops distance from the specific central peer. *Nbr(A i, hop t ): The neighbour set of Ai, and the shortest distance between peers in the set and A i is exact t hops. *Unnecessary message: The repeated flooding query message received by a peer.

Rules for Predicting Unnecessary Messages A i is father of A j, that is to say A j has received message from A i. A j will send the message to its neighbours (A k ) except A i. Then A j can predict that other peers may also send the messages to A k, define these peers (including A j ) as Possible Father Set (PFS) of A k, expressed as : PFS(A i, A j, A k )={Nbr(A k, hop 1 ) Nbr(A i, hop 1 )}

Unnecessary Message prediction Rules Rule 1: if A k Nbr(A i, hop 1 ), then Msg(A j, A k ) must be unnecessary message. Rule 2: if A k Nbr(A i, hop 1 ), and PFS(A i, A j, A k ) ≠ {A j }, then Msg(A j, A k ) is possible unnecessary message, or else it is necessary message.

Proposed Neighbour Information Based Routing Algorithm Step1: If A i has received the query before, drop the query. Step2: If the query is satisfied, then return results to sponsor. Step3: If TTL is not more than zero, drop the query.

Proposed Neighbour Information Based Routing Algorithm (Contd...) Step4: For each A j Nbr(A i, hop 1 ), predicts the validity of Msg(A i,A j ) by UMP. If it is a necessary message, then send it to A j ; if it is unnecessary message; do not send it; if it is a “possible” unnecessary message judged by UMP Rule2 then do not send it.

Advantages of The Proposed Approach This algorithm does not reduce the search scope of the general flooding algorithm. The node utilizes only its local NIT. No wastage of time in exchanging information between nodes to decide the valid neighbours. Retrieving information from local database is around five times faster than exchanging topology information

Conclusion Our simulations suggest that unnecessary messages make up to 27% in different kinds of network topologies algorithm attains an efficiency of 73% of necessary messages.