1. Efficient Retrieval of User Contents in MANETs Δημόκας Νικόλαος Data Engineering Laboratory, Aristotle University of Thessaloniki

2. MANETs Nodes are free to move, join or leave the network Bandwidth constrained links Multi-hop communication P2P approach in most cases Limited computation power and cache space Two fundamentals problems arise: How to discover services and resources available at other nodes How to transfer information between two network nodes

3. Eureka Key idea: Exploit the information density concept and allow users to estimate where in MANET the information they are looking for can be found Advantages: Waste of bandwidth is avoided by selectively forwarding content queries Fewer replies messages Fewer collisions The use of GPS is not required Applicable for VANETs: The road topology reduces the regions where information can be found Highly mobile environment

4. System and Assumptions (1/2) One or more gateways Each node is equipped with a data cache Pure P2P system N data items. Each item is divided into units, called chunks The missing chunks can be retrieved from different nodes Each node requests a data item i with rate λ i. Each node knows the number of chunks into which a data item is divided A node responds with information message A node rebroadcasting a request stores and sets query status to pending

5. System and Assumptions (2/2) All nodes listen to channel => A pending query could be transformed to solved Each query header includes a HOP_COUNT Each node computes an information density estimate in a distributed manner for each item Requester node adds to the header the ESTIMATED_DENSITY Node forwards a request if its own estimate is higher than that carried by the request It is used a query lag time similar to DSR It is used a query time to live to shorten the reach of broadcast queries

6. Information Density Estimation (1/4) The information density function, δ i (x, y), is defined as the spatial density of information chunks cached at nodes participating in the network, around a point whose spatial coordinates are (x, y) At each sampling step j, a node n computes an information density sample s i,j (n) for each data item it is aware of, by using information captured within its reach range The sample s i,j (n) represents the estimated number of new copies of chunks (during step j) Local information density sample s l i,j (n) If node n generates a reply message to node Q the contribution that is added to s l i,j (n) is: 1-(h Q -1)/TTL

7. Information Density Estimation (2/4) If node n receives a new transiting information message from node I to node Q, the contribution is: (1-(h Q -1)/TTL)+(1-(h I -1)/TTL) The last case accounts for the reception of an information message whose contribution must not (or cannot) be related to a corresponding query the contribution is: 1-(h I -1)/TTL

8. Information Density Estimation (3/4) Distributed information density sample s d i,j (n) Every time a node m generates or relays a query for some chunks of data item i, it advertises its local information density sample for this item, s l i,j (m) A node n receiving the query computes: ∑ mєMi,j (n) s l i,j (n)/|Mi,j(n)| Overall information density sample For each step j : (s l i,j (n) + s d i,j (n)) / 2

9. Information Density Estimation (4/4) Filtering and information density estimate The filter is built so that the value of each new sample is kept almost constant to its original value for W sampling steps since it was computed, after which it is exponentially decreased The average cache time of chunks at nodes is 1/µ The sampling frequency is given by: f c = 1/ T c = Wµ The larger the W, the higher the sampling frequency