High performance Big Data Systems for spatial, spatio-temporal and graph data management

点击翻译按钮获取中文摘要

Big Data is a transformative technological trend that is poised to change how businesses operate, governments function, and people work. The explosive growth of data is fueled by applications like Twitter and Facebook, and the proliferation of devices, mobiles phones, satellite imagery, software logs, cameras, and RFID readers. Traditional database techniques are inadequate to deal with this data deluge. It is necessary to devise fundamentally new approaches to data management, tools and techniques.This research proposal addresses the key challenges of Big Data: volume, velocity, variety and complexity. Many spatio-temporal applications, such as Location-Based Services (LBS), are characterized by high frequency of location data updates received from numerous GPS-enabled devices. Traditional relational databases are unable to sustain the high update and query throughput typical in LBS. To address this, my research will develop a scalable and high throughput distributed main-memory data store, capable of supporting millions of location updates and thousands of queries per second. The proposed system will exploit fast in-memory processing over high speed networks.Spatial data, stored in Geographic Information Systems (GIS) and satellite imagery, are characterized by volume, variety and complexity of processing. Existing database approaches either do not support parallel spatial query processing or cannot deal with its variety: vector, raster and microdata formats. As a solution, I am building a parallel database system for heterogeneous spatial Big Data.The real value of Big Data comes from its potential to help companies improve operations and make faster, more intelligent decisions. It is important to improve the productivity of software developers and reduce the cost of IT operations. To that end, my research will conduct a thorough re-examination of the complete stack of Big Data processing frameworks, with a view to develop an efficient analytics framework with an intuitive programming model.Actionable insights derived from data analytics can be critically important in healthcare services. Current data management approaches to Electronic Medical Records (EMR) fall short when it comes to improve therapeutic decision making and treatment. To address this issue, this research will develop a semantic graph database for healthcare analytics. The idea is to improve the ability to retrieve medical records that are “similar to” a present case and thus it can serve as a valuable resource to improve the quality of decision.This research has strong potential to provide economic and social benefits to Canada. The improved capability to handle data, with high volume, velocity and complexity, will impact various industries from LBS, energy, transportation to agriculture and healthcare. Moreover, this research with offer training to HQP in how to deal with the challenges of Big Data.