Research On Discovery Service Of IOT Based On BTLChord

The Internet Of Things(IOT) is booming up worldwide, and more and more related researches and applications are emerging. The query of item information is an important part of IOT applications. A lot of item information is generated and stored during the manufacture and distribution process in many EPCISs of companies in the supply chain which may have no direct relationship between each other. Discovery Service(DS) must be used to find all the information related to one item. DS becomes one of the most important parts of IOT.There's no standard specification about DS till Now. Previous studies prove that DS based on DHT-P2 P has excellent performance. DHT-DS includes Single layer DHT-DS and semi-distributed DHT-DS. Semi-distributed DHT-DS solves the problem about logic location and the physical location mismatch and does better in route efficiency than the single layer DHT-DS, but the dependence on the super node makes the extensibility worse. System with different overlay network algorithms which are Chord and 4 kinds of improved Chord are studied and simulated. We found that there is still a problem about load balancing because of the heterogeneous network nodes in the network and non-uniform distribution of query. Some nodes are overload but some nodes are idle at the same time.To solve the above problem, BTLChord is proposed in this paper. BTLChord includes LI(Load Information), EFT(Enhanced Finger Table) and DUP(DUPlicate). Load Information is added to all the nodes in the finger table in LI. The next hop is chosen by considering two factors, distance and load information, in routing process. LI improves the load balance but weakens the route efficiency. The backup super node will work cooperatively with heavy load super node in the same group to solve the overload issue caused by too much accessing load. Neighbor nodes' information are added to all the nodes in the finger table in EFT. LI+EFT will have more candidate nodes than LI and result in better balance and route efficiency. DUP is only implemented in ordinary nodes. By overloaded node resources copies distributed to other lightly loaded nodes to ease access to ordinary nodes overload problem, and further improve the load balance.Simulations are performed in PeerSim in our simulation environment. The results proved that BTLChord is 30% to 37% better than existing semi-distributed DHT-DS based on ABDChord in routing efficiency, 30% in query success rate, but the routing information storage space grows 4.8 times. With the same group query rate increasing, BTLChord is better than single layer DHT-DS from 73.7% to 93.7% in routing efficiency. At the same time, the query success rate of BTLChord goes up from 73.7% to 93.7% and the single layer DHT-DS based on ABDChord goes down from 98.9% to 93.0%. BTLChord's routing information storage space grows 1.2 times.