orchestration

Background: Network slicing has turned out to be one of the key enablers in the 5G networks due to the ability to support the diverse applications such as ultra reliable and low latency communications for the self-driving cars or IoT-like massive machine type communications. Prior expeditions lacked integrated tools for the dynamic assignment and allocation of resources and no possibility for maintaining constant QoS. Objective: In this article, the primary aim is to synthesis and test a reinforcement learning–driven slicing framework in order to orchestrate the resources of the three types of slices – URLLC, mMTC, and eMBB. This is to improve the performance of the sliced resource, ensure high availability, and minimize competition of the resources in multi-tenant scenarios in 5G networks. Methods: The proposed study design includes a focus on the key stakeholders and their needs for requirements gathering and an experimental field for actual implementation. Resource distribution is guided by the reinforcement learning algorithms by trying to minimize a cost function which incorporates the relation between the latency, isolation, throughput and energy expended. Using a number of runs, quality of performance is monitored to enable assessment of stability as well as response rates. Results: Experimental results show that the proposed framework achieves a lower level of latency violations and capacity oversubscription compared to heuristic methods. Furthermore, it consistently achieves nearly 2.5X better throughput for telemedicine slices and guarantees less than 5 ms latency for time-sensitive services during dynamic traffic conditions. Conclusion: The study shows how reinforcement learning can be effective and applied for end-to-end 5G network slicing. This sort of adaptive orchestration can increase service dependability while optimising overhead and herald instantly climbable multi-tenant networks compatible with various industries