QoS-Aware Approximate Query Processing for Smart Cities Spatial Data Streams.

Area of Science:

• Data Science
• Computer Science
• Geospatial Data Analysis

Background:

• Increasing volumes of georeferenced data from IoT devices present challenges for stream processing systems.
• Mobility data's skewed nature and fluctuating arrival rates complicate achieving latency and accuracy goals.
• Existing systems struggle to balance real-time processing demands with data quality requirements.

Purpose of the Study:

• To propose ApproxSSPS, a novel system for approximate processing of geo-referenced mobility data streams at scale.
• To provide quality of service guarantees for stateful aggregations and top-N queries on spatial data streams.
• To develop a system that interactively learns latency statistics and dynamically adjusts sampling rates.

Main Methods:

• Developed ApproxSSPS, a system featuring a controller for interactive learning of latency statistics.
• Implemented dynamic sampling rate calculation to meet specified latency and accuracy targets.
• Evaluated ApproxSSPS on Apache Spark Structured Streaming using real-world mobility data.

Main Results:

• ApproxSSPS successfully meets latency and accuracy targets across diverse parameter configurations and load intensities.
• The system demonstrates a significant performance improvement over a state-of-the-art baseline system.
• ApproxSSPS effectively balances latency and accuracy, providing a plausible trade-off.

Conclusions:

• ApproxSSPS is a novel spatial data stream processing system capable of delivering accurate results in a timely manner.
• The system's dynamic data sampling approach ensures quality of service for mobility data streams.
• ApproxSSPS offers a robust solution for handling the complexities of large-scale, real-time geospatial data.