Abstract
One of the major challenges that network researchers and operators face today is the lack of reliable and scalable network testbeds. Since it is often infeasible to perform experiments directly on a production network or build analytical models for complex systems, researchers often resort to simulation or downscaled testbed experiments. However, designing a downscaled experiment that can faithfully represent a large-scale experiment is often challenging. The results of a non-representative experiment can be misleading and unexpected bugs may not be discovered until the Internet protocol or application is deployed into an operational network. In this work, we present two solutions to enable large-scale network experiments. Our first solution, flow-based scenario partitioning (FSP), is a platform-independent mechanism to partition a large network experiment into a set of small experiments that are sequentially executed. Each of the small experiments can be conducted on a given number of experimental nodes, e.g., the available machines on a testbed. Results from the small experiments approximate the results that would have been obtained from the original large experiment. Experimental results from several simulation and testbed experiments demonstrate that our techniques approximate performance characteristics, even with closed-loop traffic and congested links. Our second solution, EasyScale, aims to bridge the current gap between emulation testbed users and large-scale security experiments possibly using multiple scaling techniques. EasyScale is a new framework for easily configuring a large-scale network security experiment on an emulation testbed. Multiple scaling techniques, such as full and OS-level virtualization techniques, can be used for different parts of the input experimental topology in order to balance scalability and fidelity. The EasyScale resource allocation scheme considers user-specified fidelity requirements. Additional resources are allocated to the experiment components that are considered to be highly important, in order to increase the experimental fidelity. Our results from distributed denial of service and worm attack experiments demonstrate that EasyScale can easily allocate testbed resources to the critical components in an experiment, lowering the barrier for testbed users to conduct high fidelity yet scalable network security experiments.^
