Lee, Hee Won. Network Emulation with Adaptive Time Dilation. (under the Direction of Network Emulation with Adaptive Time Dilation

LEE, HEE WON. Network Emulation with Adaptive Time Dilation. (Under the direction of Mihail L. Sichitiu and David Thuente.) In order to evaluate the performance of new network architectures and protocols, researchers have used diverse methods such as theoretical analysis, network simulation, testbed implementation, and emulation. Among these methods, emulation has been widely employed over the past decade because it is more flexible than simulation in that it does not need to re-implement the protocols while maintaining much of the realism of testbeds. Since virtual machines (VMs) allow a single physical host (PH) to create multiple virtual nodes, it is possible to construct an emulation testbed for large-scale networked systems with a relatively small number of PHs. However, as a PH is heavily loaded with multiple virtual nodes, emulation accuracy decreases due to the execution delay. Time dilation allows VMs to perceive higher performance than that of their physical machine. We present an approach using adaptive time dilation to emulate large-scale networks composed of heterogeneous operating systems (OSs). Our emulation system dynamically adapts virtual time rates to system and network loads created by the virtual nodes. To evaluate the system, distributed VMs running Linux, Windows, FreeBSD, and Junos are emulated on general purpose physical machines. Our system also adapts the degree of time dilation to emulate communication links of arbitrary latency and data rate. To evaluate this system, virtual hosts (VHs) distributed over multiple PHs are heavily loaded by applications that generate UDP and TCP traffic. Finally, we build an integrated emulation and simulation environment based on adaptive time dilation, which significantly improves the accuracy of the integrated system without the use of adaptive time dilation. © Copyright 2015 by Hee Won Lee All Rights Reserved Network Emulation with Adaptive Time Dilation