Leveraging the Cloud for Integrated Network Experimentation

The goal of this research is to determine the feasibility of performing integrated network experimentation using cloud services. This research uses performance metrics to compare computing architectures constructed in the cloud to architectures that run on traditional networks. If so, then cloud network architectures will display the same expected behavior as traditional network architectures, thus allowing the construction of networking testbeds at potentially substantial cost savings. Since the Amazon cloud does not support broadcast or multicast traffic, distributed applications face a challenge. Many distributed applications use broadcast or multicast to communicate real-time information. This research includes a case study for developing a distributed network application in the cloud which overcomes the restriction on broadcast and multicast traffic. During performance testing, the baseline network and cloud network configurations are provided statistically equivalent traffic workload. Metrics such as packet loss, delay, jitter and throughput are compared to determine relative performance. Analysis of the experimental results shows that in each case, the cloud network configurations performed at or above the performance level of the baseline network. Therefore, the public cloud infrastructure is suitable for performing integrated network experimentation. This research continues Project Everest\u27s efforts to leverage cloud services for network experimentation. Project Everest is a framework which aims to combine emulation and cloud infrastructure into a single testbed using the Amazon Elastic Compute Cloud (EC2). Their tests indicate satisfactory cloud performance, but they recommend testing cloud network performance under various workload. This research carries out those performance tests