RGCA: A Reliable GPU Cluster Architecture for Large-Scale Internet of Things Computing Based on Effective...

Area of Science:

• Computer Science
• Parallel Computing
• Internet of Things

Background:

• Internet of Things (IoT) environments generate large-scale data requiring efficient processing.
• Existing high-performance computing solutions may not be cost-effective or reliable for diverse IoT data characteristics.

Purpose of the Study:

• To develop a low-cost, high-performance, and high-reliability computing system for IoT data mining.
• To enhance IoT services through optimized data processing on GPU clusters.

Main Methods:

• Developed an Energy Consumption Calculation Method (ECCM) for Wireless Sensor Networks (WSNs).
• Proposed a Two-level Parallel Optimization Model (TLPOM) using CUDA to optimize GPU resource allocation (blocks and threads).
• Dynamically coupled Thread-Level Parallelism (TLP) and Instruction-Level Parallelism (ILP) for performance gains without increased energy consumption.
• Integrated ECCM and TLPOM into a Reliable GPU Cluster Architecture (RGCA) for high-reliability computing.

Main Results:

• The TLPOM achieved average performance increases of 34.1% (Fermi), 33.96% (Kepler), and 24.07% (Maxwell).
• The RGCA ensured low-cost and high-reliability services for the IoT computing system.
• Optimized resource planning and compiler techniques improved algorithm performance within node constraints.

Conclusions:

• The proposed system offers a cost-effective and reliable solution for large-scale IoT data processing.
• The combination of ECCM, TLPOM, and RGCA significantly boosts computational performance and system dependability for IoT applications.