Middleware infrastructure for processing of big spatial data on Spark

Location-based services such as Recon Instruments’ Engage web application provide feedback to users based on analysis of GPS trajectory data. While existing data analysis platforms such as Hadoop- GIS provide simplified support for batch processing of massive amounts of “big spatial data”, they are not efficient for supporting iterative machine learning algorithms (such as those in use at Recon) or interactive queries (common in web applications). More recently, a newer class of cluster-based solutions, known as “in-memory” databases, such as Apache Spark [3], has been introduced to address scalability for these types of workloads. The goal of this project is to provide the same easy to use interface as traditional spatial analysis platforms but leveraging Spark’s cluster-based in-memory scalability over multiple nodes. Recon will benefit from both (i) research into the application of in-memory databases to GIS analysis and (ii) the delivery of a software prototype for improving the support of Apache Spark for GIS tasks.

Faculty Supervisor:

Dr. Eric Wohlstadter

Student:

Reza Babanezhad Harikandeh

Partner:

Recon Instruments Inc.

Discipline:

Computer science

Sector:

Digital media

University:

University of British Columbia