The rise of high performance computing has seen a shift of services from locally managed Data Centers, to centralized globally redundant Data Centers (Cloud Computing). The scale of operation and churn required for cloud computing has in turn led to the rise of faster and programmable network pathing, via SDN & NFV. Cloud compute resources are accessible to individual researchers, as well as larger organizations. Cloud computing relies heavily on virtualization and abstraction of resources. The interconnect between these resources is more complex than ever, due to the need to seamlessly move from virtual to physical to hybrid networks and resources. MPLS as a technology is robust and has been used as transport for decades with a good track record. QoS has been available within most protocols to ensure service levels are maintained. The integration of MPLS, QoS and virtual environments is a space of increasing interest. It would allow for the seamless movement of traffic from end to end without the need for specialized hardware or vendor lock-in.
In this thesis, the performance gains of IP/MPLS networks utilizing QoS on commercially available virtual environments has been investigated and studied. Latency was captured via round trip time metrics and tabulated for voice, video and data, with QoS and congestion as the primary differentiators. The study discusses the approach taken, the common thinking, and finally analyzes the results of a simulation, in order to show that MPLS & QoS benefits are viable in virtualized environments.
Networking and System Administration (MS)
Madhavan, Amar, "MPLS & QoS in Virtual Environments" (2017). Thesis. Rochester Institute of Technology. Accessed from