Novel Adaptive Multi-User Multi-Services Scheduling to Enhance Throughput in 5G-Advanced and Beyond

Sharvari Ravindran,Saptarshi Chaudhuri,Jyotsna Bapat,Debabrata Das

DOI: https://doi.org/10.1109/tnsm.2024.3351669

2024-01-01

IEEE Transactions on Network and Service Management

Abstract:Optimum network slice resource allocation for multiple User Equipment (UE) requesting several services concurrently in a scalable manner is a complex open research problem. Each network slice is designed to support distinct applications with corresponding Service Level Agreements (SLAs) across multiple UEs. The distribution of network resources across slices has been a challenge to achieve the SLAs under limited bandwidth and power budget constraints. This is referred to as inter-slice resource allocation, where a UE may simultaneously request multiple services hosted across slices. When a UE is assigned a slice, the slice shares the available resources across new and existing UEs without violating the SLA. This is referred to as intra-slice scalability. While inter-slice resource allocation has been addressed in the literature, further throughput gains can be achieved by improving intra-slice scalability. This paper presents a novel Probabilistic Intra-slice Resource Service Scheduling (PRSS) method. PRSS algorithm works in two stages. In the first stage, the service throughput is estimated using a multinomial probabilistic model, followed by dynamic conditional resource estimation in the second stage. For newly requested services, the resources are re-estimated and reconfigured across prior inter and intra-slice services. The proposed two-stage method allows the system to be highly adaptive with respect to the number of UEs and their requested services hosted across slices. Analytical and simulation results show that by using PRSS, the number of services served with a better experience in terms of service quality, i.e., throughput is enhanced by 20% to 75% compared to state-of-the-art schedulers.

computer science, information systems

What problem does this paper attempt to address?

The paper aims to address the issue of network slicing resource allocation when multiple user equipment (UE) concurrently request various services in 5G and post-5G networks. Specifically, the paper focuses on how to optimize resource allocation within network slices under limited bandwidth and power budget constraints to improve Quality of Service (QoS), particularly throughput.

The paper proposes a new Probabilistic Intra-Slice Resource Service Scheduling method (PRSS) to achieve this goal through a two-stage process:
1. The first stage uses a polynomial probability model to estimate service throughput.
2. The second stage performs dynamic conditional resource estimation.

For new service requests, PRSS re-estimates and reconfigures previously allocated intra- and inter-slice service resources. This approach allows the system to dynamically adjust based on the number of UEs and their requested services, thereby significantly enhancing system adaptability and throughput. Experimental results show that, compared to existing state-of-the-art schedulers, using the PRSS algorithm can improve QoS and throughput by 20% to 75%.