Benchmarking Microservice-based Application
Modern distributed computing applications have extremely complex structures. Such structural complexity poses significant challenges when these applications are looking to leverage edge computing, which provides low-latency and high-throughput computing proximal to end users. This project addresses the performance issue of large-scale distributed applications by employing an application-network co-design approach.
Continuously evolving research plan.
Environment Setup
Setup microservice-based demo application.
Distributed Tracing
Trace the demo application’s network traffic across different components.
Data Modeling
Building statistical/machine-learning models to show the relationship among key factors and performance metrics.
Project Evolution
Design the next research challenge to tackle.
Wireless Environment
A wireless network environment is set up including a Kubernete Server Cluster, a wireless router, and several Raspberry Pis used to simulate clients.
Multi-dimensional Data
We have collected network-tracing data with respect to different setups and present them here interactively.
Download Dataset
Uncertainty-agnostic Data Models
We build state-of-the-art data models to fit the collected network traces.
More
Uncertainty-aware Data Models
We also trained Uncertainty-aware Probabilistic Neural Net (PNN) models to fit the collected network traces with confidence intervals.
More