Data-driven Bandwidth Adaptation for Radio Access Network Slices
Authors:
Panagiotis Nikolaidis,
Asim Zoulkarni,
John Baras
Abstract:
The need to satisfy the QoS requirements of multiple network slices deployed at the same base station poses a major challenge to network operators. The problem becomes even harder when the desired QoS involves packet delays. In that case, network utility maximization is not directly applicable since the utilities of the slices are unknown. As a result, most related works learn online the utilities…
▽ More
The need to satisfy the QoS requirements of multiple network slices deployed at the same base station poses a major challenge to network operators. The problem becomes even harder when the desired QoS involves packet delays. In that case, network utility maximization is not directly applicable since the utilities of the slices are unknown. As a result, most related works learn online the utilities of all slices and how to split the resources among them. Unfortunately, this approach does not scale well for many slices. Instead, it is needed to perform learning separately for each slice. To this end, we develop a bandwidth demand estimator; a network function that periodically receives as input the traffic of the slice and outputs the amount of bandwidth that its MAC scheduler needs to deliver the desired QoS. We develop the bandwidth demand estimator for QoS involving packet delay metrics based on a model-based reinforcement learning algorithm. We implement the algorithm on a cellular testbed and conduct experiments with time-varying traffic loads. Results show that the algorithm delivers the desired QoS but with significantly less bandwidth than non-adaptive approaches and other baseline online learning algorithms.
△ Less
Submitted 28 April, 2024; v1 submitted 28 November, 2023;
originally announced November 2023.
Resource Efficiency vs Performance Isolation Tradeoff in Network Slicing
Authors:
Panagiotis Nikolaidis,
Asim Zoulkarni,
John Baras
Abstract:
We consider the tradeoff between resource efficiency and performance isolation that emerges when multiplexing the resource demands of Network Slices (NSs). On the one hand, multiplexing allows the use of idle resources, which increases resource efficiency. On the other hand, the performance of each NS becomes susceptible to traffic surges in other NSs, which degrades performance isolation. The ana…
▽ More
We consider the tradeoff between resource efficiency and performance isolation that emerges when multiplexing the resource demands of Network Slices (NSs). On the one hand, multiplexing allows the use of idle resources, which increases resource efficiency. On the other hand, the performance of each NS becomes susceptible to traffic surges in other NSs, which degrades performance isolation. The analysis of this tradeoff enables network operators to determine the effect of performance isolation on the operating cost of each NS.
To study the tradeoff, we solve an optimization problem where we find the multiplexing policy that requires the least provisioned resources to honor the Service Level Agreements (SLAs) of all NSs. The SLA of each NS i states that its resource demand should be met for $P^H_i$ fraction of time, and for $P^L_i \leq P^H_i$ fraction of time, it should be met regardless of the demands of other NSs.
For resource demands that follow ergodic Markov chains, we show that the well-known Max-Weight scheduler is an optimal multiplexing policy. Since the Max-Weight scheduler does not require any knowledge of the statistics of the resource demands, we also propose its use in non-markovian settings. For resource demands obtained in the LTE module of ns-3, we show that the Max-Weight scheduler reduces the provisioned bandwidth by 36.2% when no performance isolation is required. Lastly, for these non-markovian resource demands, the Max-Weight scheduler maintains its optimality since it requires as much provisioned bandwidth as the best non-causal scheduler.
△ Less
Submitted 26 March, 2023;
originally announced March 2023.
