Abstract:Network function virtualization allows the implementation of network services on readily available commercial off-the-shelf servers. A primary obstacle is efficiently, autonomously, and dynamically allocating physical resources to virtual network functions that experience fluctuating resource demands. Many existing optimization approaches typically assume static resource demands for each virtual network function instance. Nonetheless, this may result in either resource waste or compromised service quality when an excessive or insufficient allocation of resources occurs. Effectively overseeing virtual network functions presents a considerable challenge due to their dynamic nature. The allocation of resources should be adjusted to align with fluctuations in incoming network traffic. These variations cause significant delays in resource reallocation. Therefore, employing resource estimation models prior to allocation can preemptively address potential issues, fostering dynamic performance enhancements in resource allocation methods. The proposed work utilizes several ensemble machine learning techniques, such as adaboost, bagging, extratrees, histgradient boosting, and lightGBM, and ML models such as bayesian, lasso and poisson regression models, for accurately predicting resource requirements and analyzing performance using CPU resources and throughput. A matching theory-based virtual network function placement algorithm is used for resource allocation. The simulation results show that the acceptance ratio of the ensemble machine learning models integrated with the matching theory-based placement approach is four times higher than that of the fixed resource allocation model. Also, the average delay of the fixed model is five times higher than that of the proposed resource prediction and allocation model. Simulation results also show that the performance of the ensemble machine learning models is better when compared to standard ML models in terms of SFC acceptance ratio and total delay.

Reservation of Virtualized Resources with Optimistic Online Learning

Online VNF Chaining and Predictive Scheduling: Optimality and Trade-offs

Adaptive Resource Allocation for Virtualized Base Stations in O-RAN with Online Learning

Online Optimization for Network Resource Allocation and Comparison with Reinforcement Learning Techniques

Resource Reservation Within Sliced 5G Networks: A Cost-Reduction Strategy for Service Providers.

Efficient Function Placement in Virtual Networks: An Online Learning Approach

Exponentially Weighted Algorithm for Online Network Resource Allocation with Long-Term Constraints

Online Optimization for Randomized Network Resource Allocation with Long-Term Constraints

Online Allocation with Replenishable Budgets: Worst Case and Beyond

Learning to Cover: Online Learning and Optimization with Irreversible Decisions

Stochastic Averaging for Constrained Optimization With Application to Online Resource Allocation

Dynamic Distributed Edge Resource Provisioning Via Online Learning Across Timescales.

Optimistically Tempered Online Learning

Learning-Augmented Decentralized Online Convex Optimization in Networks

Active Learning for Fair and Stable Online Allocations

Computing Optimal Regularizers for Online Linear Optimization

Model-Based Reinforcement Learning Framework of Online Network Resource Allocation

Online Learning in Decentralized Multiuser Resource Sharing Problems

Learning-based Two-tiered Online Optimization of Region-wide Datacenter Resource Allocation

The Power of Online Learning in Stochastic Network Optimization.

Predictive resource allocation and VNF deployment using ensemble learning