In developing countries such as Senegal, information-sharing systems are mainly faced to the lack of IT resources: low internet connection speed, limited storage resources, etc. To overcome these shortcomings, an alternative is P2P networks usage.
In P2P Information retrieval systems, the main challenges are the bandwidth consumption and the network latency management.
Currently we are working on FreeCore, a distributed hash table (DHT) based scheme that provide efficient keywords-based search services.
The main contribution of FreeCore is the construction of an efficient over-DHT index that resolve keywords-search problems.
This problem has been widely addressed in the past and several approaches have been proposed in order to solve it, but any of them does really do it.
First of all, there are a group of solutions proposing over-DHT based schemes that use inverted lists as data structure. The advantage of these solutions is they keep the properties offered by DHT such as the scaling, the efficient routing protocoles and the high data availability. But they suffer for high bandwidth consumtion and high search latence that are inherent for inverted lists intersections.
On the other hand, there are some other solutions building ad hoc structure indexes or modifying the DHT overlay. These solutions generally use content summary as data structure (Bloom Filters, binary vector, etc.) which have the advantage of reducing indexing costs and bandwidth consumtion but suffer for the high latence needed to solve keyword-search.
To solve these unacceptable problems, FreeCore combines these two principles. So FreeCore represents documents and queries by Bloom filter (FB) and maintains an over-DHT index that permits to perform BF supersets searches.
Intuitively, a Bloom filter A is a superset of a Bloom filter B, if both FBs have the same number of bits, and all bits set to 1 in B are set to 1 in A.
Thus, to resolve a query, we simply build its FB and then recover FB documents that are supersets of the FB of the request.