An online cost minimization of the slice broker based on deep reinforcement learning

The fifth generation of mobile networks can provide differentiated services enabled by network slicing and multi-access edge computing. A new business actor, called Slice Broker (SB), is emerging as an intermediate entity that buys networking and computing resources from Infrastructure Providers (InPs) and provides network slices to Slice Tenants (STs). This paper addresses the problem of online jointly allocating all the network slices requested by the STs and selecting the resources to purchase from the InPs. The target of the problem is the minimization of the SB costs. We assume that a network slice is implemented as a sequence of virtual network functions and that the InPs sell the networking and computing resources on predetermined configurations. The addressed problem is solved by using deep reinforcement learning with a model -free policy gradient -based algorithm. The proposed solution is evaluated and compared with benchmark solutions in various scenarios. The results show that the benchmark solutions have a cost that is from 10% to more than twice higher than the proposed solution in all scenarios.