Advances, Systems and Applications
Table 12 An overview of the various load-balancing solutions for SD-IoT
From: An overview of QoS-aware load balancing techniques in SDN-based IoT networks
Disadvantages | Advantages | Key Contribution | Main Subject | Conference/Journal |
|---|---|---|---|---|
- Not considering other aspects of QoS such as security, capacity - Lack of evaluation of energy consumption |  + Delay-sensitive task processing  + Improved delay and QoS | Cloud/fog network architecture | Improved delay for real-time service processing | China Communications (IEEE) [33] |
- Inefficiency of load-balancing scheme for saturation scenarios - Not using virtualization in FoT-Gateway |  + Reduced response time and lost samples | Programming to select a virtual machine | Load-balancing for FoT -Gateways and network links | International Conference on Internet of Things (IEEE) [30] |
- Starvation in medium and low-priority applications |  + Acceptance control to ensure QoS of high-priority applications  + Load-balancing between the routes and selection of the route with the maximum bandwidth  + Reduced delay, jitter, and packet loss  + Improved average end-to-end flow performance | Admission control | Application- aware QoS routing | Symposium on Computers and Communications (IEEE) [14] |
- Lack of cost management |  + Improving resource efficiency and response time  + Considering the types of services (service classification) | Type of service request | Load-balancing among cloud servers | IEEE Communications Magazine [89] |
- Increased overhead at the data layer with frequent rerouting - Lack of attention to other criteria such as security - Dependence on the transfer rate - Maintenance of backup paths |  + Improved response time  + Being used in human/machine networks  + Reducing communication overhead | Traffic-aware load-balancing | Improved QoS by detecting and rerouting traffic | IEEE Internet of Things Journal [32] |
- Single-point of failure - The lack of evaluation of other criteria of QoS such as congestion, overload, and security |  + Improved end-to-end delays and packet delivery rates  + Reliable, scalable, and secure communication network | Traffic routing optimization | Global load- balanced routing the problem in the AMI network | IEEE Internet of Things Journal [99] |
- Single-point of failure |  + Reduced delay and ensured safe network execution  + Improved network security and stability | Deploying the middlebox in the right place | SDN-based data transfer security model in IoT based on middlebox | IEEE Internet of Things Journal [79] |
- Lack of evaluation of energy consumption |  + Timely identification of attack models  + Network scale support | Network partitioning and fog resource allocation | Large-scale intrusion detection with minimal delay | IEEE Access [2] |
- The lack of evaluation of energy consumption and carbon emissions |  + Improved delay, throughput, and resource efficiency  + Improved network performance | Data transfer architecture | Management of communication resources | IEEE Internet of Things Journal [80] |
- Checking other parameters of QoS such as security and energy consumption |  + Improved response time, resource efficiency | Vertical (hierarchical) structure of the controller pool | Large-scale control layer load-balancing | IEEE Access [20] |
- Single-point of failure - Not using machine learning at max-overload - Not Considering heterogeneous resources |  + Improved response time, cost, resource utilization, and energy consumption  + Increase task acceptance rate | Workload tolerance | QoS-aware load balancing | IEEE Access [95] |
- Considering the incentive mechanism for well-behaved devices |  + Reduce task completion time  + Resource utilization at edge devices | Token-based resource management | Efficient resource allocation of edge nodes | IEEE Sensors Journal [85] |
-The migration process leads to increased network delay - Suitable for a limited number of target controllers to choose from - The lack of attention to the heterogeneity of tasks and resources - The high cost of migration for a large-scale environment (migration overhead) |  + Faster achievement of load-balancing at the control layer  + Lower communication overhead and reduced response time | Multi-criteria decision making | Load-balancing in the control plane | IEEE Internet of Things Journal [37] |
- Lack of conscious mechanisms for load-balancing of servers - Lack of QoS management at the layer of distributed SDN control and multi-domain network - Requires Network Function virtualization (NFV) for energy management and QoS |  + IoT traffic classification  + Scalability of IoT infrastructure with maintaining QoS  + Achieving justice and reducing the impact of corruption in QoS  + Increasing throughput, and resource efficiency | Resource and QoS-aware framework | Scalable traffic management | IEEE Internet of Things Journal [38] |
-Controller bottleneck used - The lack of privacy protection - Need to predict malicious activity with ML techniques - High migration overhead - Low scalability |  + Increased load-balancing and optimal use of resources  + Increased security  + Improved response time, packet delivery rate, delay, throughput, and overhead | Secure edge computing framework | Lightweight authentication scheme | IEEE Access [26] |
- Lack of server and network integration using virtualization techniques - Not considering large-scale networks - Need for processing requests based on priority and resource allocation - Not evaluating traffic classification to ensure QoS |  + Minimization of the bandwidth costs  + Link and server load-balancing  + Considering load-balancing at network and server levels  + Consider homogeneous and heterogeneous networks  + Suitable for evaluating any fog computation topology | Cooperative Fog-Cloud Computing Architecture | Load-balancing to manage resources | IEEE Access [11] |
- Lack of extensive control of wireless parameters - Controller bottleneck used |  + Improved packet loss rate, received signal strength, and throughput  + Reduced dependence on the controller  + No controller overload | QoS-aware load-balancing | Solving network congestion problems based on the load level | IEEE Access [44] |
- Using queues and their effect on delay - The lack of evaluation of energy consumption |  + Improved response time and reliability  + Accelerated user access to sensor data | Load-balancing based on multi-criteria decision-making | Achieving load fairness and reducing service processing delays | IEEE Internet of Things Journal [28] |
- Using queues and their effect on delay - No cost analysis - The lack of evaluation of energy consumption |  + System stability in high current input fluctuations  + Ensuring fairness in resource allocation | Cloud-edge hierarchical system | Increased scalability and reduced computational delay | IEEE Systems Journal [45] |
- Single point of failure controller |  + Avoid congestion and E2E delay  + QoS guarantee, improving resource efficiency  + Overhead reduction | Traffic engineering framework | Resource management among slices | IEEE Network [100] |
- Non-consideration of other aspects of QoS such as scalability, network lifetime, and energy consumption |  + Reduced data redundancy and service response delay  + Mobility support | Cloud/edge computing | Service synchronization and data aggregation | IEEE Internet of Things Journal [5] |
- Non-consideration of QoS criteria |  + Mobility management, handover optimization  + Improved scalability | Distributed hash-based monitoring structure | Flow control and mobility management in heterogeneous urban networks | IEEE Transactions on Parallel and Distributed Systems [9] |
- Increased delay in providing almost optimal routing solutions - The lack of appropriate algorithms for traffic forecasting - Non-consideration of effective network performance parameters |  + Improved load-balancing | Approximate routing algorithms | Routing optimization problem with TCAM capacity constraint | Journal of Communications and Networks (IEEE) [48] |
- Using queues and their effects on delay - The lack of focus on resource efficiency - Need for scalability improvement |  + Minimizing queues, request processing time, and balancing the controller load  + Reduce immigration costs | Multi-objective optimization | Self-Adaptive Load-Balancing | International Conference on Autonomic Computing and Self-Organizing Systems (IEEE) [52] |
- Lack of improved switching efficiency among IoV services in fog clusters |  + Four-objective optimization  + Minimum delay and energy consumption  + Maximum load-balancing and service stability  + Mobility support  + Using heterogeneous computational resources  + Improving real-time scalability | Architecture based on cloud-fog computing | Resource allocation in fog clusters | IEEE Transactions on Intelligent Transportation Systems [16] |
- Increased energy consumption due to handover functions |  + Mobility support  + Improved load-balancing, service response, and handover rates  + Reduced congestion and increased service availability  + Considering a heterogeneous network | Link assignment | Load-balancing at the control layer | 14th International Conference on Communication Systems & Networks (IEEE) [92] |
- Data redundancy in neighbouring tables sent to the controller - Lack of QoS management in the mode of distributed control |  + Reduced the number of messages  + Reduced energy consumption  + Prolong the network's lifetime | Load-balancing-based routing and clustering | Reduced load distribution and increased network lifetime | IEEE Access [67] |
- Non-anticipation of QoS criteria with artificial intelligence techniques - Inattention to scalability - Non-examination of the heterogeneity of tasks and resources |  + Improved throughput, response time, and resource efficiency  + Maximum CPU usage and minimum memory usage  + Checking the migration cost and load-balancing rate | QoS -aware load-balancing framework | Improved QoS for network stability | IEEE Transactions on Green Communications and Networking [90] |
- Further investigation to reduce the response time of the controller when a failure occurs - Non-consideration of the large scale |  + Increase link utilization, balance traffic loads, conserve table space  + Reduce blocked packets, and alleviate table-full events | Reroute traffic flows | Load-balancing between links of switches | IEEE Transactions on Network and Service Management [81] |
- Not considering controller overhead - Not using machine learning in a multi-controller scenario |  + Minimizing the impact of link failure  + Better performance for delay-sensitive services  + Improved throughput, energy consumption, delay | Efficient and reliable routing | Reliability-aware flows distribution | IEEE Transactions on Vehicular Technology [82] |
- Data analysis of nodes with cloud technologies - Considering algorithms to compatible with 5G infrastructure |  + Throughput, delay, packet loss rate  + Support wireless communication protocols  + Time-sensitive prioritization | machine learning-based load-balancing | Distribution of nodes to base stations | IEEE Internet of Things Journal [88] |
- Single point of failure controller |  + Improved throughput, round-trip delay, packet loss rate | Scheduling to calculate rerouting | Load balance of link traffic | International Conference on Measuring Technology and Mechatronics Automation (IEEE) [97] |
- Non-consideration of the packet processing priority - Controller bottleneck - Non-use of a combination of transmission paths for optimization of load-balancing - Testing non-extremity of fixed pockets/non-fixed pockets |  + Classification of tasks by type of service  + Improved data transfer time and load-balancing  + Optimal local prevention | Service-Oriented SDN-SFC | Programming data transfer routes | Journal of Network and Computer Applications (Elsevier) [42] |
- The need to minimize the cost of fulfilling requests |  + Improving throughput and load-balancing  + Considering communication delays and calculations  + Maximum acceptance of requests | Cloudlet network framework on the mobile edge | Resource management and load-balancing | Future Generation Computer Systems (Elsevier) [7] |
- Need for achieving complete network control among fog nodes with data layer Programming |  + End-to-end routing  + Reliable (bandwidth guarantee)  + Improved throughput and response time  + Efficient for large-scale systems  + Increased system availability  + Reduced delay in finding the offloading node | Dynamic offloading service between fog nodes | Finding the optimal node to handle tasks | Future Generation Computer Systems (Elsevier) [29] |
- Non-implementation of network traffic based on real-world applications |  + Improved packet delivery rate, packet loss, and delay | Admission control | Network flow management and congestion reduction | Computer Networks (Elsevier) [101] |
- Higher communication overhead - The lack of identity and prevention of security attacks - The need for load balance between heterogeneous devices |  + Reduced rotation and waiting time  + Improved processing performance and use of network resources | Hierarchical architecture of controllers | Network management and load-balancing among devices | Journal of Network and Computer Applications (Elsevier) [22] |
- The lack of resource efficiency |  + Reduced delay and energy consumption  + Solving resource pricing problems between the user and the edge resource provider | Energy-aware resource allocation | Improved QoS in edge computing | Sustainable Computing: Informatics and Systems (Elsevier) [41] |
- The lack of evaluation of energy consumption, network lifetime, and packet delivery rate |  + Meeting scalability and delay requirements  + Improved response time, packet loss rate, and processing time | SDN network programming | Load-balancing for the Fog of Things Platforms | Journal of King Saud University—Computer and Information Sciences (Elsevier) [17] |
- Migration overhead |  + Prevent control plane overhead and distribute traffic efficiently  + Reduce response time and cost of migration | Dynamic switch migration | Load-balancing among controllers | Computer Networks (Elsevier) [91] |
- Need for high privacy in a decentralized model - Achieving online task offloading and resource allocation with cooperating massive IoT networks |  + Improved reliability, delay, energy  + Privacy-preserving, security, and confidentiality by blockchain  + Higher throughput and lower overhead | Blockchain-based Deep Reinforcement Learning | Energy-aware task scheduling and offloading | Future Generation Computer Systems (Elsevier) [96] |
- Combining the presented approach with security-aware scheduling approaches |  + Improved load-balancing, delay  + Meeting the security requirements of IoT devices  + Reduce response time | Security-aware workflow scheduler | Joint security and performance optimization | Journal of Information Security and Applications (Elsevier) [98] |
- Need for optimization algorithms to load-balancing at the data plane - Need for hybrid machine learning algorithms for packet analysis |  + Improved bandwidth, response time, delay, and packet loss  + Considering security metrics such as detection accuracy and authentication time | Using honeypots, blockchains, and vSwitches | Providing secure multi-controller load-balancing | Future Generation Computer Systems (Elsevier) [94] |
- Extend on dynamic network |  + Optimizing packet delivery ratio, average latency, network lifetime, and energy consumption | Traffic flow optimization | Energy efficient routing | Sustainable Energy Technologies and Assessments (Elsevier) [84] |
- Extending the proposed framework to a more large-scale SDN - Non-compliance of distributed architecture with security frameworks |  + Optimization of migration time, response time, and controller load  + Improved CPU usage, latency, communication cost, and throughput | Switch migration | Multi-domain SDN slave controller load balancing | Journal of King Saud University—Computer and Information Sciences (Elsevier) [43] |
- Need to apply machine learning techniques - Non-implementation of Fog and Edge computing |  + Increased security  + Improved throughput, delay, response time, and resource utilization  + Improved the durability, stability, and load balancing | Blockchain-SDN-based secure architecture | Traffic load management of real-time applications | Digital Communications and Networks (Elsevier) [15] |
- Single-point failure controller used |  + Reduced concerns about resource scarcity  + Network congestion elimination  + Improved delay, resource efficiency, and throughput  + Less number of handovers | Data offloading and load-balancing | Reduced short-term resource shortages and network congestion | Journal on Wireless Communications and Networking (Springer) [49] |
- Non-consideration of other aspects of QoS - Need for implementation of the algorithm in the real SDC-FN platform - The lack of evaluation of energy consumption |  + Mobility support  + Improved delay and response time | Cloud / Fog network architecture | Reduced real-time service delay | International Conference on Communication and Networking in China (Springer) [86] |
- Need for practical application and performance analysis - Interaction of unauthorized users with each other - Fault to check fault tolerance |  + Improved delay and throughput  + Improved load-balancing, scalability, accessibility, integrity, and network security  + Heterogeneity support | Virtualization of network functioning | SDN-based distributed IoT network | Cyber Security and Computer Science (Springer) [10] |
- Increase transfer time, and packet loss rate |  + Improved response time and Throughput | Load-balancing optimization | Load distribution between SDN controllers in IoT application | Wireless Personal Communications (Springer) [36] |
- Failure points of switches and controllers - Delaying the load-balancing function with multiple migrations - High cost of migration in a large-scale environment |  + Increased response time, resource efficiency  + Improved fault tolerance and reliability for migration | Monitoring and classification of the service | SDN-based load-balancing service | Wireless Personal Communications (Springer) [25] |
- Improper management of multiple attacks - Need for the deployment of distributed blockchain technology for confidential data management and security |  + Dealing with the epidemic damage of the Covid-19 virus in the industry  + Ensuring security and reliability  + Improved throughput, response time, and packet loss rate | SDN-based IoT architecture with NFV | Productivity of industry potentials in the Covid-19 pandemic | Cluster Computing (Springer) [3] |
- Evaluation of load-balancing and traffic-based decisions for green cloud computing |  + Improved throughput, bandwidth utilization, response time | Machine Learning for routing and server selection | Load-balancing in DCN Servers | Arabian Journal for Science and Engineering (Springer) [83] |
- The need for scheduling with the load-balancing of flight nodes |  + Improved throughput, packet delivery rates, and end-to-end delays  + Increased network lifetime and traffic balancing | Computational load distribution between nodes | Distributed traffic congestion control | Electronics (MDPI) [65] |
- Unstable performance - Need for checking other goals, such as reliability - Need for other searching criteria in the optimization algorithm - Non-evaluation of the selection of non-dominated solutions based on angle or distance |  + Improved energy consumption, cost, and run time | Using multi-objective optimization | Load-balancing in cloud computing | Sensors (MDPI) [6] |
- Need to expand the security parameters and more performance - Not testing the proposed technique in a real test-bed environment | - Improved Response time, energy consumption, and communication delay | Secure and energy-aware fog computing architecture | Load-balancing to improve utilization of resources | Sustainability (MDPI) [87] |
- Increased transfer delay - Lack of evaluation of energy consumption |  + Improved E2E delay, resource efficiency  + Achieving a fair allocation of resources  + Maximization of profitability of service providers  + Improved Quality of Experience (QoE) | Hierarchical architecture of controllers | Assigning requests to cloud data centers | Multimedia Systems Conference (ACM) [47] |
- Single point failure central controller - Starvation in tasks with lower priority - Challenges the cloud for long distances with the user |  + Improved response time, throughput  + Assigning CPU resources to high-priority tasks | Task classification | Load-balancing in cloud network links | Workshop on Advanced Research and Technology in Industry Applications (Atlantis Press) [77] |
- Inefficient use of resources - Non-consideration of QoS |  + Minimization of the overall cost of communication | Controller placement based on clustering | Load-balancing between multiple controllers | Scalable Computing [56] |
- Super-controller bottleneck - Non-consideration of migration costs and the distance between controllers and switches in overload controllers - Non-consideration of resource efficiency |  + Reduced delay  + Improved load-balancing | Real-time delay-based load-balancing | Simultaneous overload of multiple controllers | Computers, Materials & Continua (Tech Science Press) [12] |
- Not testing the proposed strategy in real scenarios - Non-consideration of other performance criteria such as energy consumption and response time |  + Improved load-balancing,  + Reduced number of controllers and average delay and delay | Controller placement | Load-balancing and reduced packet release delay | Computers, Materials & Continua (Tech Science Press) [102] |
- Non-evaluation of energy consumption, loss rate, and packet delivery |  + Minimized delays  + Reduced completion time of tasks  + Potential for mobility and location-awareness | Cloud/edge computing architecture | Improved load-balancing and performance in latencies | Conference Proceedings (AIP) [93] |