Efficiency Analysis of High-dimensional Matrix Convolution and Multiplication Based on Serverless Platform

With the rapid evolution of cloud-native, more and more cloud services are becoming Serverless, which has become a major trend in the future of cloud computing. Non-Serverless cloud services that are billed according to reserved resources can already be classified as traditional techniques with lots of disadvantages. The utility of serverless and its difference from traditional computing frameworks is explored by performing high-dimensional matrix operations with serverless technology provided by cloud providers. In this paper, the efficiency and infect factors of high-dimensional matrix multiplication and convolution are validated. It is found that the execution time and computation time of functions vary with matrix size, CPU, and convolution kernel when performing matrix multiplication and convolution operations. The parallel operation is significantly more efficient when performing matrix multiplication. The efficiency of serverless operations can be optimized by selecting the appropriate CPU and other variables and decomposing tasks for parallel execution to avoid resource competition.