1-1 CMPE 259 Sensor Networks Katia Obraczka Winter 2005 Storage and Querying

1-2 Overview

1-3 Overview r Sensors sense/generate data. r Users/applications interested in data. r Require an infrastructure for data access and storage. r Common user operations are: m Queries (monitoring). m Actuate and control.

1-4 Types of queries r Historical: m What is the average rainfall over past 2 days? r Current: m What is the current temperate in Rm# 226? r Long running: m Temperature in rm# 226 over the next 4 hours every 30 seconds.

1-5 Issues r How to identify relevant sensors? r Computation vs. communication tradeoff. m Where to process query? Inside the sensor network (route query). At centralized location (route data). –Large amounts of data transfer: efficiency? m How to process query?

1-6 DataSpace: Querying and Monitoring Deeply Networked Collections in Physical Space T. Imielinski and S. Goel, Rutgers University r Billions of objects populate space. r Each produces and locally stores data. r Location aware. r Can be selectively monitored, queried and controlled.  Physical world enhanced with data.

1-7 Characteristics r Dataspace m Data lives on the object m Users access not only “local” information but can navigate entire dataspace m Spatial world divided in 3-D datacubes CS Bldg., street, block etc m Communication, messaging and computation techniques for querying and monitoring required

1-8 Querying and monitoring r Queries are spatially driven. r Steps: m Identify relevant datacubes. m Identify relevant nodes (dataflocks). Datacube directory service. m Aggregation for queries on several datacubes. E.g.,information about Manhattan taxi cabs

1-9 Architecting DataSpace r Querying and monitoring. m Multicast mechanisms. (attribute, value) pair mapped to a multicast address. m Group membership based on: Physical location. Attribute (temperature, #vehicles, etc.). m Mapping (attribute, value) to multicast address. E.g., using hash tables. m Condition on the index attribute of query mapped to multicast address. Query reaches all objects satisfying condition on index attribute. Objects perform check for remaining conditions and respond or not.

1-10 Where to provide functionality? r Network versus application layer.

1-11 Network as DataSpace engine Space Handle identifies datacube. Subject Handles are (attribute, value) pairs such that attribute is indexed. DataSpace address Based on location of datacube Based on relevant attribute

1-12 Using geographic routing infrastruture r Route message based on physical location. m Use GPS coordinates for location. r Once query reaches a region use mutlicast.