Abstract

Fifth-generation (5G) mobile networks are introducing revolutionary changes in the design, deployment, and management of cellular infrastructure. Key technologies such as network slicing, which is unprecedented in the mobile communication domain, serve as one of the key corners of the 5G architecture. Each slice is individually customized to accommodate various applications, including Enhanced Mobile Broadband (eMBB), Ultra-Reliable Low-Latency Communications (URLLC), and massive Machine-Type Communications (mMTC). To the best of our knowledge, this paper presents the first detailed investigation of NS performance in 5 G Core (5GC) networks, along with a comprehensive study of architecture, deployment models, and empirical performance metrics in both real and emulated conditions.
The primary objective of this work is to quantify the efficiency with which network slicing can achieve quality of service guarantees and fulfill SLAs across various service domains. We examine how slicing performs in terms of core network isolation, latency control, throughput stability, and scalability under multi-tenant, dynamic, and service-differentiated workloads. To achieve this, the study utilizes Software-Defined Networking (SDN) and Network Function Virtualization (NFV) as building blocks to create a programmable, elastic network infrastructure that enables slice orchestration and lifecycle management. We design, instantiate, and monitor individual slices using a service-based architecture (SBA), and analyse their performance in terms of KPIs such as RTT, PLR, jitter, and CPU/memory usage.
To simulate practical deployment, the evaluation utilizes a virtualized 5 G Core (5GC) environment with open-source simulators, including Open5GS, Mininet, and ONOS. The core of the research presents controlled traffic emulations for various use case categories on these platforms. For eMBB, we consider high-throughput data traffic, like 4K Video Streaming and File download scenarios. For URLLC, we consider latency-sensitive applications, such as remote surgery and Autonomous vehicle control applications. For mMTC, we consider the massive amount of IoT sensor data used to model ultra-dense, low-rate transmissions. These controlled settings enable comparison of slice-level resource isolation, congestion behavior, and response to orchestration changes under different load profiles in the network.
Preliminary results indicate that network slicing yields significant improvements in resource utilization and service differentiation, particularly when combined with dynamic slice scaling and policy-based resource allocation. eMBB slices maintain consistent throughput with little packet drop, even under bursty traffic. When using dedicated resource pools and preemptive scheduling approaches, URLLC slices exhibit latency of less than 10 ms. mMTC slices demonstrate resilience in supporting tens of thousands of simultaneous low-bandwidth devices with low orchestration latency. However, there were also various shortcomings, including inter-slice interference when resource boundaries are not rigorously isolated, higher control plane latency resulting from slice instantiation delay, and the need for intelligent orchestration to support CSI scaling and healing.
This paper makes several interesting contributions. First, it provides a measurable evaluation of 5G network slicing performance based on open and reproducible testbeds. Second, it performs a trade-off analysis between slice isolation and resource efficiency. Third, it also furnishes comparative program benchmarks for the three 5G service categories in support of SLA control policies. Ultimately, it provides a roadmap for improving slice orchestration, which involves integrating AI-based policy engines and distributed edge deployments.
The performance analysis presented in this paper confirms the practicality of network slicing in 5G core networks for delivering fine-grained, QoS-compliant services. However, to make the best use of it, the next step will focus on optimizing the orchestration layer, enhancing resource isolation, and integrating predictive analytics to adapt slices proactively. These enhancements are crucial for meeting emerging 5G use cases and enabling scalable, long-term support for evolving vertical markets.