Workload allocation in edge computing deals with distributing computational tasks or workloads among different nodes in an edge computing network. Edge computing has multiple tiers with different computing power. Workload allocation is vital in determining the tiers to handle the workload or number of tasks assigned to an individual part.
The goal of workload allocation is to optimize the use of resources and ensure that the computing tasks are processed efficiently and effectively. Edge computing involves processing data and providing services closer to the network-s edge, such as at the device level rather than in a central data center.
The allocation of workloads between cloudlets for each application influences the response time of application requests, taking into detail both network latency and computational latency. It enables faster processing times, improved reliability, and reduced latency. Therefore, two other workload allocation resources are present :
• Density-based clustering (DBC) strategy: The idea of DBC is to offload user workloads to suitable cloudlets until the workloads of the cloudlets exceed the average workload among cloudlets.
• Latency-based strategy: The latency-based strategy minimizes the network delay between Apps and cloudlets by assigning apps to suitable cloudlets.
Based on allocation, it has two major types of allocating the workloads,
• Static allocation: This involves assigning tasks to edge nodes based on predefined rules and configurations without considering the current status of the network or the nodes.
• Dynamic allocation: Includes assigning tasks to edge nodes based on real-time information about the status of the network and the available resources at each node. Dynamic allocation can result in more efficient use of resources and improved performance.
Workload allocation in edge computing is a complex and challenging task including :
• Latency requirements: Some applications require low latency, and workload allocation algorithms must be able to meet these requirements by executing critical tasks on edge devices.
• Resource constraints: Edge devices have limited computing and storage resources, and workload allocation algorithms must consider this while deciding which tasks should be executed locally and which should be offloaded to the cloud.
• Real-time processing: Edge computing systems often require real-time processing of data, and workload allocation algorithms must be able to make decisions in real-time to minimize latency.
• Network conditions: Edge computing systems operate in unpredictable network conditions, and workload allocation algorithms must adapt to these conditions in real-time to minimize latency and improve the system-s overall performance.
• Network-aware algorithms: Future research could focus on developing algorithms aware of the network conditions and can dynamically adjust the workload allocation to minimize latency and improve performance.
• Real-time algorithms: Future research could focus on developing algorithms that can make decisions about workload allocation in real-time, taking into account the current network conditions and resource utilization of the devices.
• Integration with other technologies: Future research could focus on integrating workload allocation algorithms with other technologies, such as 5G networks and IoT devices, on improving the overall performance of edge computing systems.
• Optimal Workload Allocation for Edge Computing Network Using Application Prediction
• Workload Re-Allocation for Edge Computing with Server Collaboration.
• A Dynamic Deep Neural Network Design for Efficient Workload Allocation in Edge Computing.
• Intelligent workload allocation in IoT–Fog–cloud architecture towards mobile edge computing.
