Invited Talk: High-Performance Modeling and Simulation of Computer Networks

High-Performance Modeling and Simulation of Computer Networks

May 26, 2017

Department of Computer Science
Tsinghua University, Beijing, China
Host: Professor Zhihui Du (都志辉)

Abstract: Modeling and simulation (M&S) plays an important role in the design analysis and performance evaluation of complex systems. Many of these systems, such as computer networks, involve a large number of interrelated components and processes. Complex behaviors emerge as these components and processes inter-operate across multiple scales at various granularities. M&S must be able to provide sufficiently accurate results while coping with the scale and complexity.

My talk will focus on two novel techniques in high-performance network modeling and simulation. The first is a GPU-assisted hybrid network traffic modeling method. The hybrid approach offloads the computationally intensive bulk traffic calculations to the background onto GPU, while leaving detailed simulation of network transactions in the foreground on CPU. Our experiments show that the CPU-GPU hybrid approach can achieve significant performance improvement over the CPU-only approach.

The second technique is a distributed network emulation method based on simulation symbiosis. Mininet is a container-based emulation environment that can study networks consisted of virtual hosts and OpenFlow-enabled virtual switches on Linux. It is well-known, however, that experiments using Mininet may lose fidelity for large-scale networks and heavy traffic load. The proposed symbiotic approach uses an abstract network model to coordinate distributed Mininet instances with superimposed traffic to represent large-scale network scenarios.

Invited Talk: High-Performance Modeling and Simulation of Computer Networks

High-Performance Modeling and Simulation of Computer Networks

April 26, 2017

Laboratory of Information, Networking and Communication Sciences (LINCS), Paris, France
Host: Professor Dario Rossi

Abstract: Modeling and simulation (M&S) plays an important role in the design analysis and performance evaluation of complex systems. Many of these systems, such as computer networks, involve a large number of interrelated components and processes. Complex behaviors emerge as these components and processes inter-operate across multiple scales at various granularities. M&S must be able to provide sufficiently accurate results while coping with the scale and complexity.
My talk will focus on two novel techniques in high-performance network modeling and simulation. The first is a GPU-assisted hybrid network traffic modeling method. The hybrid approach offloads the computationally intensive bulk traffic calculations to the background onto GPU, while leaving detailed simulation of network transactions in the foreground on CPU. Our experiments show that the CPU-GPU hybrid approach can achieve significant performance improvement over the CPU-only approach.
The second technique is a distributed network emulation method based on simulation symbiosis. Mininet is a container-based emulation environment that can study networks consisted of virtual hosts and OpenFlow-enabled virtual switches on Linux. It is well-known, however, that experiments using Mininet may lose fidelity for large-scale networks and heavy traffic load. The proposed symbiotic approach uses an abstract network model to coordinate distributed Mininet instances with superimposed traffic to represent large-scale network scenarios.

Invited Talk: Symbiotic Modeling and High-Performance Simulation

Symbiotic Modeling and High-Performance Simulation

January 19, 2017

Department of Computer Science, Colorado School of Mines
Host: Professor Tracy Camp

Abstract: Modeling and simulation plays an important role in the design analysis and performance evaluation of complex systems. Many of these systems, such as the internet and high-performance computing systems, involve a huge number of interrelated components and processes. Complex behaviors emerge as these components and processes inter-operate across multiple scales at various granularities. Modeling and simulation must be able to provide sufficiently accurate results while coping with the scale and the complexity of these systems. My talk will focus on some of our latest advances in high-performance modeling and simulation techniques. I will focus on two specific case studies, one on network emulation and the other on high-performance computing (HPC) modeling.
In the first case, I will present a novel distributed network emulation mechanism based on modeling symbiosis. Mininet is a container-based emulation environment that can study networks consisted of virtual hosts and OpenFlow-enabled virtual switches on Linux. It is well-known, however, that experiments using Mininet may lose fidelity for large-scale networks and heavy traffic load. We propose a symbiotic approach, where an abstract network model is used to coordinate the distributed emulation instances superimposed to represent the target network. In doing so, we can effectively study the behavior of real implementation of network applications on large-scale networks in a distributed environment.
In the second case, I will present our latest work on performance modeling of HPC architectures and applications. In collaboration with the Los Alamos National Laboratory, we have developed a highly efficient simulator, called Performance Prediction Toolkit (PPT), which can facilitate rapid and accurate performance prediction of large-scale scientific applications on existing and future HPC architectures.

TOMACS’15 Paper: Cluster-Based Spatiotemporal Background Traffic

Cluster-Based Spatiotemporal Background Traffic Generation for Network Simulation, Ting Li and Jason Liu. ACM Transactions on Modeling and Computer Simulation (TOMACS), 25(1), Article No. 4, January 2015. [paper]

abstractbibtex
To reduce the computational complexity of large-scale network simulation, one needs to distinguish foreground traffic generated by the target applications one intends to study from background traffic that represents the bulk of the network traffic generated by other applications. Background traffic competes with foreground traffic for network resources and consequently plays an important role in determining the behavior of network applications. Existing background traffic models either operate only at coarse time granularity or focus only on individual links. There is little insight on how to meaningfully apply realistic background traffic over the entire network. In this article, we propose a method for generating background traffic with spatial and temporal characteristics observed from real traffic traces. We apply data clustering techniques to describe the behavior of end hosts as a function of multidimensional attributes and group them into distinct classes, and then map the classes to simulated routers so that we can generate traffic in accordance with the cluster-level statistics. The proposed traffic generator makes no assumption on the target network topology. It is also capable of scaling the generated traffic so that the traffic intensity can be varied accordingly in order to test applications under different and yet realistic network conditions. Experiments show that our method is able to generate traffic that maintains the same spatial and temporal characteristics as in the observed traffic traces.
@article{Li2014:bgtraffic,
author = {Li, Ting and Liu, Jason},
title = {Cluster-Based Spatiotemporal Background Traffic Generation for Network Simulation},
journal = {ACM Trans. Model. Comput. Simul.},
issue_date = {January 2015},
volume = {25},
number = {1},
month = nov,
year = {2014},
issn = {1049-3301},
pages = {4:1–4:25},
articleno = {4},
numpages = {25},
url = {http://doi.acm.org/10.1145/2667222},
doi = {10.1145/2667222},
acmid = {2667222},
publisher = {ACM},
address = {New York, NY, USA},
}