Dynamic Resource Management in Service-Oriented Core Networks (Wireless Networks)

✍ Scribed by Weihua Zhuang, Kaige Qu

• Publisher: Springer
• Year: 2021
• Tongue: English
• Leaves: 182
• Category: Library

✦ Synopsis

This book provides a timely and comprehensive study of dynamic resource management for network slicing in service-oriented fifth-generation (5G) and beyond core networks. This includes the perspective of developing efficient computation resource provisioning and scheduling solutions to guarantee consistent service performance in terms of end-to-end (E2E) data delivery delay.
Network slicing is enabled by the software defined networking (SDN) and network function virtualization (NFV) paradigms. For a network slice with a target traffic load, the E2E service delivery is enabled by virtual network function (VNF) placement and traffic routing with static resource allocations. When data traffic enters the network, the traffic load is dynamic and can deviate from the target value, potentially leading to QoS performance degradation and network congestion. Data traffic has dynamics in different time granularities. For example, the traffic statistics (e.g., mean and variance) can be non-stationary and experience significant changes in a coarse time granularity, which are usually predictable. Within a long time duration with stationary traffic statistics, there are traffic dynamics in small timescales, which are usually highly bursty and unpredictable. To provide continuous QoS performance guarantee and ensure efficient and fair operation of the network slices over time, it is essential to develop dynamic resource management schemes for the embedded services in the presence of traffic dynamics during virtual network operation. Queueing theory is used in system modeling, and different techniques including optimization and machine learning are applied to solving the dynamic resource management problems.
Based on a simplified M/M/1 queueing model with Poisson traffic arrivals, an optimization model for flow migration is presented to accommodate the large-timescale changes in the average traffic rates with average E2E delay guarantee, while addressing a trade-off between load balancing and flow migration overhead. To overcome the limitations of Poisson traffic model, the authors present a machine learning approach for dynamic VNF resource scaling and migration. The new solution captures the inherent traffic patterns in a real-world traffic trace with non-stationary traffic statistics in large timescale, predicts resource demands for VNF resource scaling, and triggers adaptive VNF migration decision making, to achieve load balancing, migration cost reduction, and resource overloading penalty suppression in the long run. Both supervised and unsupervised machine learning tools are investigated for dynamic resource management. To accommodate the traffic dynamics in small time granularities, the authors present a dynamic VNF scheduling scheme to coordinate the scheduling among VNFs of multiple services, which achieves network utility maximization with delay guarantee for each service. Researchers and graduate students working in the areas of electrical engineering, computing engineering and computer science will find this book useful as a reference or secondary text. Professionals in industry seeking solutions to dynamic resource management for 5G and beyond networks will also want to purchase this book.

✦ Table of Contents

Preface
Contents
Acronyms
1 Introduction
1.1 Service-Oriented Core Networks
1.1.1 Network Slicing Framework
1.1.1.1 Infrastructure Domain
1.1.1.2 Tenant Domain
1.1.1.3 SDN-NFV Integration
1.2 Motivation and Research Topics
1.2.1 Dynamic Flow Migration
1.2.2 Dynamic VNF Resource Scaling and Migration
1.2.3 Dynamic VNF Scheduling
1.3 Outline
References
2 Literature Review
2.1 Service Function Chain (SFC) Embedding
2.2 Elastic SFC Provisioning
2.2.1 Reconfiguration Overhead Awareness
2.2.2 QoS Awareness
2.2.3 Resource Scaling Triggers
2.2.4 Data-Driven Adaptivity
2.3 Dynamic VNF Scheduling
2.4 Summary
References
3 Dynamic Flow Migration: A Model-Based Optimization Approach
3.1 System Model
3.1.1 Services
3.1.2 Virtual Resource Pool
3.1.3 Computing Model
3.1.3.1 M/M/1 Queueing Model
3.1.3.2 Processing Density Model
3.1.3.3 Multi-VNF Computing Resource Sharing
3.1.4 Elastic SFC Provisioning
3.1.4.1 Joint VNF Migration and Vertical Scaling
3.1.4.2 Flexible Virtual Link Provisioning
3.1.4.3 VNF State Transfer
3.2 Optimization Model for Dynamic Flow Migration
3.2.1 Constraints
3.2.1.1 VNF Migration: VNF to NFV Node Re-mapping
3.2.1.2 Resource Scaling
3.2.1.3 Joint VNF Migration and Resource Scaling
3.2.1.4 Subflow to Virtual Link Re-mapping
3.2.1.5 VNF State Transfer
3.2.2 Optimization Problem
3.3 Optimal MIQCP Solution
3.3.1 Constraint Transformation
3.3.2 Optimality Gap and Optimum Mapping
3.4 Low-Complexity Heuristic Flow Migration Algorithm
3.4.1 Algorithm Overview
3.4.1.1 Migration Cost Reduction: Alternate Resource Scaling and Migration
3.4.1.2 Load Balancing: Alternate Resource Scaling and Threshold Update
3.4.2 Redistribution of Hop Delay Bounds
3.4.2.1 Step I: Delay Scaling for SFC Category III
3.4.2.2 Step II: Delay Scaling for SFC Category II
3.4.3 Loop 1: Sequential Migration Decision
3.4.4 Loop 2: Iterative Resource Utilization Threshold Update
3.4.5 Complexity Analysis
3.5 Simulation Results
3.6 Summary
References
4 Dynamic VNF Resource Scaling and Migration: A Machine Learning Approach
4.1 System Model
4.1.1 Network Model
4.1.2 Traffic Model
4.1.2.1 Multi-Timescale Time Series
4.1.2.2 Stationary Traffic Segments with Unknown Change Points
4.1.2.3 Factional Brownian Motion for Each Stationary Traffic Segment
4.1.3 Resource Provisioning Model
4.2 Resource Demand Prediction for Non-stationary Traffic
4.2.1 Bayesian Online Change Point Detection
4.2.1.1 Run Length
4.2.1.2 Posterior Run Length Distribution
4.2.1.3 From Stochastic Run Length Distribution to Deterministic Change Points
4.2.2 Traffic Parameter Learning
4.2.2.1 A Look-Back Scheme for Traffic Sample Collection
4.2.2.2 Fractional Brownian Motion Traffic Model
4.2.2.3 Gaussian Process Regression
4.2.3 Resource Demand Prediction for a Stationary Traffic Segment
4.3 Dynamic VNF Resource Scaling and Migration Framework
4.4 Dynamic VNF Migration Decision
4.4.1 Markov Decision Process
4.4.1.1 Action
4.4.1.2 State
4.4.1.3 Reward
4.4.2 Deep Reinforcement Learning
4.4.2.1 Deep Q-Learning
4.4.2.2 Penalty-Aware Prioritized Experience Replay
4.5 Performance Evaluation
4.5.1 Simulation System Setup
4.5.2 Simulation Results
4.5.2.1 Change Point Detection
4.5.2.2 Traffic Parameter Learning
4.5.2.3 Resource Demand Prediction
4.5.2.4 VNF Migration
4.6 Summary
References
5 Dynamic VNF Scheduling for Network Utility Maximization
5.1 System Model
5.1.1 Services
5.1.2 Network Model
5.1.3 Discrete-Time VNF Packet Processing Queueing Model
5.1.3.1 Physical Packet Processing Queues
5.1.3.2 Delay-Aware Virtual Packet Processing Queues
5.1.4 Per-VNF FCFS Prioritized Packet Processing
5.2 Stochastic VNF Scheduling
5.2.1 Scheduling Problem Formulation
5.2.2 Lyapunov Optimization and Problem Transformation
5.3 Online Distributed Algorithm
5.3.1 Auxiliary Variable Decision
5.3.2 VNF Scheduling Decision
5.3.3 Queue Updates
5.3.4 Performance Optimality
5.4 Performance Evaluation
5.5 Summary
References
6 Conclusions and Future Research Directions
6.1 Conclusions
6.2 Future Research Directions
A Derivation of α̃n(k)
B Proof of Lemma 5.1
C VNF Scheduling Algorithm with Packet Rushing
Packet Rushing Analysis
Modified VNF Scheduling Algorithm