GPU-based Real-time Contact Tracing at Scale.

Area of Science:

• Computer Science
• Epidemiology
• Public Health

Background:

• Contact tracing is crucial for controlling infectious disease spread like COVID-19.
• Real-time processing of large-scale, frequently sampled tracing data presents significant computational challenges.

Purpose of the Study:

• To develop a GPU-based system for efficient, real-time contact tracing using location data.
• To address the challenges of processing massive datasets for timely disease control.

Main Methods:

• Utilized a GPU-based architecture for spatial proximity queries with temporal constraints.
• Implemented dynamic indexing of moving objects with an adaptive partitioning schema on GPU.
• Developed an efficient mechanism for evaluating and filtering valid spatial and temporal contacts.

Main Results:

• Achieved sub-second response times for contact tracing involving tens of millions of individuals.
• Demonstrated a two-magnitude performance increase compared to CPU-based approaches.
• The system effectively handles dynamic indexing and optimizes contact pair retrieval.

Conclusions:

• The proposed GPU-based system offers a highly efficient solution for real-time, large-scale contact tracing.
• This approach significantly enhances the speed and scalability of disease control efforts.
• The system's performance boost validates the use of GPU-centered parallelism for epidemiological data processing.