Network Virtualization. Of all ICT resources, the network is the least sufficiently virtualized. See for example the comparison between the state of the compute virtualization with the state of the network virtualization in today’s traditional DCN presented by Xxxxx Xxxxxxxx of Amazon in his famous 2010 talk [36]. Although there were some positive changes in this area from the 2010, adequate virtualization of network resources – on the one hand capable of providing the applications with everything they might need from the networking layer while on the other hand harnessing all the advanced and/or specialized features of the underlying communication layer – does not yet exist.
Network Virtualization. The network virtualization use case relates to the multi-tenant (cloud) environment where multiple applications owned and operated by multiple agents are simultaneously deployed in the common shared data centre infrastructure. In such environments, it is both impractical and undesirable to provide clients (application developers, operators, and users) with the physical infrastructure access (visibility and control). Instead, virtualization of the infrastructural aspects (storage, compute, and network) is mandatory in such environments to ensure proper isolation and sharing. The orchestrator’s task in the network virtualization use case is to coordinate physical resource allocation to meet the workload demand across the storage, the compute, and the network. Specifically in COSIGN, we focus on network resource management as part of scenarios requiring coordination between the storage, the compute, and the networking allocations. Such scenarios involve on-demand bandwidth provisioning to a set of use-cases, such as: back-up or storage replication processes; compute instance migration as part of disaster recovery and avoidance schemes; loading the image of the compute instance to its physical host, etc. Additional group of scenarios is related to Network Function Virtualization (NFV), where instances of network appliances have to be efficiently provisioned and managed to satisfy the policies and the constraints. In order to fulfil its role in the network virtualization use case, the orchestrator has to receive and interpret application level requests for the ICT resources. The requests can come in a form of application blue-print (in modelling language such as TOSCA), or in a form of application-level or infrastructure-level policy to be realized and maintained (e.g. disallow all communications with version Y of protocol X or ensure all inter-rack traffic is encrypted).
