Faashark: an End-to-End Network Traffic Analysis System Atop Serverless Computing

Hongyu Zhao,Shanxing Pan,Zinuo Cai,Xinglei Chen,Lingxiao Jin,Honghao Gao,Shaohua Wan,Ruhui Ma,Haibing Guan

DOI: https://doi.org/10.1109/tnse.2023.3294406

IF: 6.6

2024-01-01

IEEE Transactions on Network Science and Engineering

Abstract:The prosperity of the Internet has made network traffic analysis increasingly indispensable in network operation. With the development of machine learning, more researchers and engineers are using deep learning models for network traffic analysis. However, the rapidly growing data size and model complexity make resource scheduling a serious limitation, which is why cloud computing services are typically required for network analysis. To leverage the advantages of serverless platforms, we propose faaShark , an end-to-end network traffic analysis system based on a serverless computing platform. faaShark adapts distributed training to fully utilize the flexibility of serverless platforms. Additionally, we design a cold start optimization algorithm to reduce the hit rate of cold start when serving pretrained models to handle network analysis requests. Our extensive experiments evaluate the impact of several parameters of distributed training and confirm the effectiveness of our cold start optimization algorithm when building such a network analysis system atop serverless computing frameworks.