Enhancing the performance of the aggregated bit vector algorithm in network packet classification using GPU

Area of Science:

• Computer Science
• Network Engineering
• High-Performance Computing

Background:

• Packet classification is crucial for network functions in high-speed systems.
• Graphics Processing Units (GPUs) offer parallel processing capabilities for network tasks.
• The aggregated bit vector algorithm is suitable for parallelization.

Purpose of the Study:

• To develop and evaluate a GPU-accelerated aggregated bit vector algorithm for packet classification.
• To adapt an asymptotic analysis method for predicting the performance of the GPU kernel.
• To validate the efficiency of the GPU implementation and the accuracy of the analysis.

Main Methods:

• Implementation of a parallel kernel for the aggregated bit vector algorithm on GPUs.
• Adaptation of an asymptotic analysis technique to predict empirical results.
• Experimental evaluation of the GPU kernel's performance and analysis method's accuracy.

Main Results:

• The proposed parallel kernel demonstrates efficient packet classification on GPUs.
• The asymptotic analysis method accurately predicts experimental trends.
• GPU acceleration significantly enhances the performance of the aggregated bit vector algorithm.

Conclusions:

• The GPU-accelerated aggregated bit vector algorithm is an efficient solution for high-speed packet classification.
• The developed analysis method provides accurate performance predictions.
• This work validates the use of GPUs for accelerating network classification tasks.