The challenge of moving performance-sensitive workloads to OpenStack is all in the operations. It might be possible to show the needed performance once, in the lab, as a science experiment, but how does one take that to production across the globe on a range of different clouds? This session will describe a model-driven approach enabling reusability in 6WIND NFV operations across clouds and bare metal, using Juju for model-driven ops encapsulation and reuse, providing a case study of high performance DPDK workload operations and the VPIL VNF benchmarking service.
