Query-based wireless sensor storage management for real time applications Ravinder Tamishetty, Lek Heng Ngoh, and Pung Hung Keng Proceedings of the 2006 IEEE International Conference on Industrial Informatics (INDIN ’ 06)
Outline Introduction Location Aided data centric storage Simulation results Conclusion
Existing schemes for storage External Storage (ES) Local Storage (LS) A significant benefit of data-centric storage A group of pre-defined Low level sensor data are abstracted to high level concept of event Use a geographic hash table to map an event type into a geographic Avoid flooding
Geographic Hash Table for Data- Centric Storage (GHT) level1 mirror points root point (3,3) level2 mirror points ♦ d, hierarchy depth ♦ mirrors, 4 d -1 e.g. d = 2 (0,100) (100,0) (100,100) (0,0) The storage nodes are pre-computed and kept at the same location Keeping the storage nodes doesn ’ t consider the query space
A potential application The origin of these queries is tooted to particular region and changes periodically in the network Propose the shifting of storage node from its initial hashed location
Basic idea City Center Sensor node Storage node Query node Old storage node
Location aided data centric storage Storage node ’ s update In order to reduce the query traffic The current storage node ’ s location are not capable of keeping the data Sensor node Storage node Query node a i >r+k/2 a i <r+k/2 In the same region In the different region Storage node keeps track of the query location in a small table for a certain amount of time Query region boundary
Identify the query region boundaries In order to reduce the query traffic Sensor node Storage node Query node f: query frequency t: the waiting time for the storage node f: 4 t: 2 seconds Shirting algorithm
Shifting algorithm furthest shortest Sensor node Storage node Query node New storage node New hashing location New query region boundary identify The radius covered by region ‘ r = (d + k)/2 d: the distance between furthest and shortest query nodes from the storage node k: an additional constant is added to d as safe step Sent [c, r] to query nodes
Shifting Algorithm New storage node is identified by the hashing function v = H (key) Where key is data_type + movement Every movement of storage node the movement level is increased by one The new updated hashed location returned to the querying node and flood in the query region
Shifting Algorithm The current storage node ’ s location are not capable of keeping the data The power level at current storage node < threshold A local shifting Finds a nearest neighbor and forwards all data and they cache
Simulation results Network size: 200m*100m The number of sensor nodes: 50, 100, 200 The number of event types: 2 to 20 The number of queries: 100 to 200 The number of queries with no shift of storage node:33% The number of queries with 1 st shift of storage node:33% The number of queries with 2 nd shift of storage node:34%
Simulation results
Conclusion Presented location aided storage management Shirting algorithm Shifts the storage nodes location based on the query traffic The contributions for storage management Query region boundary estimations New storage node formations