Evaluating SPARQL Queries over Linked Data Streams

Abstract

So far we have addressed different aspects of RDF and Linked Data management, from modeling to query processing or reasoning. However, in most cases these tasks and operations are applied to static data. For streaming data, which is highly dynamic and potentially infinite, the data management paradigm is quite different, as it focuses on the evolution of data over time, rather that on storage and retrieval. Despite these differences, data streams on the Web can also benefit from the exposure of machine-readable semantic content as seen in the previous chapters. Semantic Web technologies such as RDF and SPARQL have been applied for data streams over the years, in what can be broadly called Linked Data Streams. Querying data streams is a core operation in any streaming data application. Ranging from environmental and weather station observations, to realtime patient health monitoring, the availability of data streams in our world is dramatically changing the type of applications that are being developed and made available in many domains. Many of these applications pose complex requirements regarding data management and query processing. For example, streams produced by sensors can help studying and forecasting hurricanes, to prevent natural disasters in vulnerable regions. Monitoring the barometric pressure at sea level can be combined with other wind speed measurements and satellite imaging to better predict extreme weather conditions1. Another example can be found in the health domain, where the industry has produced affordable devices that track caloric burn, blood glucose or heartbeat rates, among others, allowing live monitoring of the activity, metabolism, and sleep patterns of any person [226]. Moreover, data streams fit naturally with applications that store or publish them in the cloud, allowing ubiquitous access, aggregation, comparison,