Performance Testing for Cloud and Serverless Computing

Jan 1, 2023 · 1 min read
projects

Overview

Cloud computing introduces fundamental challenges for performance testing due to resource contention, hidden scheduling policies, and passive auto-scaling. These challenges are amplified in serverless computing environments where the resource abstraction level is higher and auto-scaling behaviors are not well-characterized.

Research Thrusts

Serverless Auto-Scaling Characterization: We define and characterize auto-scaling stages for serverless platforms, decomposing performance uncertainty into resource contention during execution and cold start-up latencies during environment initiation.

Monte Carlo Simulation-Based Testing: We develop simulation-based methodologies that can predict performance distributions of serverless applications, accounting for the stochastic nature of cloud performance fluctuations.

AI-Based Cloud Emulation: Building systems that emulate cloud environments on local machines to help users obtain accurate performance results at reduced testing costs, particularly when multiple applications need to be evaluated.

Performance Assurance in DevOps: Using development and operational data to detect performance regressions early in the software delivery cycle, with attention to end-user impact.

Cloud Computing Performance Testing Serverless DevOps
Sen He
Authors
Sen He (he/him)
Assistant Professor
Sen He is an Assistant Professor in the Department of Electrical and Computer Engineering at the University of Arizona. His research spans LLM4SE, performance engineering, computer vision, and human-computer interaction.