This paper provides a survey of resource allocation for network slicing. We focus on two classes of existing solutions: ( $i$ ) reservation-based approaches, which allocate resources on a reservation basis, and ( $ii$ ) share-based approaches, which allocate resources based on static overall shares associated to individual slices. We identify the requirements that a slice-based resource allocation mechanism should satisfy, and evaluate the performance of both approaches against these requirements. Our analysis reveals that reservation-based approaches provide a better level of isolation as well as stricter guarantees, by enabling tenants to explicitly reserve resources, but one must pay a price in terms of efficiency unless reservations can be updated very dynamically; in particular, efficiency falls below 50% when reservations are performed over long timescales. We provide further comparisons in terms of customizability, complexity, privacy and cost predictability, and discuss which approach might be more suitable depending on the network slices’ characteristics. We also describe the additional mechanisms required to implement the desired resource allocations while meeting the latency and reliability requirements of the different slice types, and outline some issues for future work.

pub