Event Details
Analyzing Run-time Performance Predictability in Software-Centric Systems Using Monte Carlo Simulation
Presenter: Zahra Nikdel
Supervisor:
Date: Fri, July 26, 2024
Time: 12:00:00 - 00:00:00
Place: Zoom, link below.
ABSTRACT
Join Zoom Meeting
https://uvic.zoom.us/j/9808571500?pwd=MlhTNFdiaTJtcG93Z2RCekJUMjFhZz09&omn=81008070246
Meeting ID: 980 857 1500
Password: dJM0y0
Note: Please log in to Zoom via SSO and your UVic Netlink ID
ABSTRACT
This research presents a detailed investigation into the run-time performance predictability of software-centric systems using Monte Carlo simulation. The focus is on assessing statistical performance measures in systems ranging from small-scale in-house deployed to large-scale cloud deployed environments, including containers and virtual machines. By introducing an OMNet++ based simulations, that incorporates various workload scenarios, OS scheduling algorithms, and deployment configurations, we aim to provide insights into the factors affecting performance predictability. Moreover, by simulating a real-world, in-production Large-scale Distributed Software System (LDSS) under statistically identical service workloads and deployment regimes, we contrast the impacts of reliable and non-reliable protocols. The results reveal that the event recovery actions triggered by reliable protocols substantially decrease the LDSS's run-time performance predictability. Key findings highlight variations in systems’ run-time behavior under similar conditions and provide practical insights for industry applications. This research contributes to a deeper understanding of performance predictability in modern software systems, offering guidance for improving the predictability and reliability of software systems.