T-NOVA announces the release of deliverable D3.3, presenting the outcomes of Task 3.3 “Service mapping”, focused on the research and the design of algorithms for the optimal mapping of network services in virtualized network infrastructures.
The deliverable proposes a review of use cases, requirements and architecture design proposed in work package 2, and an analysis of the scientific and technological state of the art in service mapping. Furthermore, the input and output interfaces for the algorithm are specified in detail, in terms of needed input data and of required output data for a service instantiation. Based on the above points, three main service mapping algorithms, together with an algorithm for service scheduling are described in the deliverable. Two service mapping algorithms are based on integer linear programming, while the third one is based on a stochastic control methodology. The mathematical formulation and the properties of the algorithms are presented and discussed in detail. Simulations results in emulated environments are proposed and discussed to highlight the characteristics on the algorithms. Prior to the description of the algorithms, a common mathematical framework for modelling the service mapping problem is proposed, based on the ETSI reference documents on virtualization architecture. The modelling framework is based on graph theory and provides a way for consistently modelling:
- The requirements of the network services to be mapped.
- The network infrastructure’s topology and its current occupancy level so that feasible and optimal mappings are computed by the algorithm.
The developed algorithms allow to consider multiple mapping objectives, including:
- Maximisation of network service requests’ acceptance.
- Minimization of mapping costs (i.e. the costs for employing the different network resources).
- Balancing of network/datacenter load distribution.
Aside of the design of the mapping algorithms, another fundamental output of this task is given by the design of a microservice-based service mapping module, aimed to host the service mapping algorithm, and of its integration inside TeNOR, the T-NOVA orchestrator. The service mapping module offers the possibility to integrate and avail of any service mapping algorithm conforming to the input/output specifications. In this way, any investigated service mapping algorithm, or even a combination of them can be potentially integrated in TeNOR, and tested.
Finally, the integration of the service mapping module including one of the developed mapping algorithms has been achieved and preliminary tested, with details reported in this document.