Abstract:In recent years, the huge expansion of Datacenters (DC) to execute billions of end-user applications in real-time leads to a large amount of energy consumption across the globe. So, the traditional TCP/IP-based networks which are being used for DC inter-connections are facing challenges of managing stringent Quality-of-Service (QoS) requirements of different applications of the end-users and service providers. Moreover, the existing solutions rely on distributed architecture and do not scale for large scale data centers. The issue of high power consumption at DC arises with the increase in the number of nodes and links in the network. Also, it becomes problematic on the DC whenever the underlying network resources (switches and routers) are not efficiently utilized at the time of peak data traffic resulting in high operational cost of energy utilization. However, Software-Defined Networking (SDN) emerges as one of the leading technologies to address the aforementioned issues using the programmable switches and controllers. Inspired from these facts, in this paper, we have formulated the Energy-Aware Routing (EAR) problem of DCs as a Mixed Integer Non-Linear Programming (MINLP) for which an Energy-Efficient Fast Flow Forwarding (EnFlow) scheme is designed. The EnFlow scheme uses the power-saving mode of the network to solve the EAR problem. It has three modules namely– priority scheduling, routing, and re-routing. The first module works according to the First-in-First-Out Push Out Priority (FIFO-POP) scheduling using the multiple OpenFlow switches. The FIFO-POP is designed to save the energy usage of multiple switches by reducing the average waiting time of incoming packets in the queue buffers. The second module is based upon an efficient flow re-routing for a new node and link adaptation to provide the maximum bandwidth to the wired links. The third module is based upon the meta-heuristic Ant Colony Routing (ACR) to execute the stochastic decision policy on the network controller for computation of the shortest path of the forwarding nodes. The proposed EnFlow scheme is simulated using the data traces of 34 cities of NorthAmerica zone with Omnet++ 5.1 using various performance evaluation metrics. The results obtained demonstrated that the proposed EnFlow scheme is 24.55% and 71.15% more energy-efficient in comparison to the RE-FPR and ILP-EAR schemes. Also, it consumes 9.72% and 40.83% lower energy in comparison to the FFHA and EXR schemes respectively.

DENA: an Intelligent Dynamic Flow Scheduling for Rate Adjustment in Green DCNs

An Efficient Sampling And Classification Approach For Flow Detection In Sdn-Based Big Data Centers

TATA: Throughput-Aware TAsk Placement in Heterogeneous Stream Processing with Deep Reinforcement Learning

Parameterized deep reinforcement learning with hybrid action space for energy efficient data center networks

Towards an energy-efficient Data Center Network based on deep reinforcement learning

Flow Splitter: A Deep Reinforcement Learning-Based Flow Scheduler for Hybrid Optical-Electrical Data Center Network.

DECC: Achieving Low Latency in Data Center Networks with Deep Reinforcement Learning

Intelligent load balancing in data center software‐defined networks

Renewable Energy-Aware Big Data Analytics in Geo-Distributed Data Centers with Reinforcement Learning

DA&FD–Deadline-Aware and Flow Duration-Based Rate Control for Mixed Flows in DCNs

Deadline-aware and Energy-Efficient Dynamic Flow Scheduling in Data Center Network.

A Modified Long Short-Term Memory-Deep Deterministic Policy Gradient-Based Scheduling Method for Active Distribution Networks

Optimal Energy System Scheduling Using A Constraint-Aware Reinforcement Learning Algorithm

Transforming Cooling Optimization for Green Data Center via Deep Reinforcement Learning

DRL-PLink: Deep Reinforcement Learning with Private Link Approach for Mix-Flow Scheduling in Software-Defined Data-Center Networks

RL-ADN: A High-Performance Deep Reinforcement Learning Environment for Optimal Energy Storage Systems Dispatch in Active Distribution Networks

EnFlow: An Energy-Efficient Fast Flow Forwarding Scheme for Software-Defined Networks

Resource Allocation in Disaggregated Data Centre Systems with Reinforcement Learning

Optimizing Data Centre Energy Efficiency via Event-Driven Deep Reinforcement Learning

RARE: Renewable Energy Aware Resource Management in Datacenters

Deep reinforcement learning based multi-layered traffic scheduling scheme in data center networks