Temporal synchronization of emulated and simulated models for assessment of networked systems

"Network experimentation practices rely on emulation or simulation methods to test new ideas. But it is difficult to exclusively simulate or emulate a broad class of emerging cyber-physical applications without incurring significant cost or loss of fidelity. This is because they involve complex interactions between cyber-entities/nodes and live physical environments over different types of networks. Analysis of such applications can be greatly simplified by constructing co-simulated models of the system where cost-effective simulations of a physical environment and the underlying network are temporally integrated with high fidelity emulations of the core application logic. In a co-simulation, certain components of the system are simulated while others are emulated. The execution of emulated and simulated components is synchronized to a shared virtual clock. The practical utility of a co-simulation platform depends on some crucial factors. Precise control over the scheduling order of emulated processes is required to make the behavior of the co-simulation independent of the CPU resources of the platform hosting it. The accuracy of ascribed virtual time for each emulated execution burst controls the rate of advancement of time perceived by the application, and the overhead associated with time synchronization between emulated and simulated components of the model affects the overall time taken for each unit of virtual time advancement. This dissertation is an attempt to tackle these challenges. We propose kernel level modifications and compiler based instrumentation of emulated code for precise instruction level tracking of process execution and control over execution order of emulated containers. This is combined with a mechanism to ascribe virtual time for each execution burst based on the sequence of executed instructions. We also discuss novel approaches to reduce virtual time synchronization overhead. This relies on compiler-based estimation of ""lookahead"", which identifies epochs of emulated execution during which a process can be predicted to act independently of any other. Through evaluations, we show that our approaches enable fast, accurate and reproducible execution of co-simulated models."

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2022

Author dc:creator

• Babu, Vignesh

Contributors dc:contributor

• Nicol, David Malcom
• Borisov, Nikita
• Caesar, Matthew
• Kalbarczyk, Zbigniew
• Liu, Jason

• Virtual Time
• Modeling
• Simulation
• Network Emulation
• Lookahead Extraction
• Virtualization
• Synchronization

Statement dc:rights

• Copyright 2021 Vignesh Babu

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/112956

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/112956