This technology enables a client node in a mobile edge computing environment to determine the computational capacity, link bandwidth, startup costs, and round-trip latency of surrogate nodes. It then calculates the workload for each node and partitions tasks into segments of varying sizes for parallel assignment. The client also processes a portion of the workload itself.
While computation offloading is a key method for supplementing resource-constrained mobile devices, varying processing performance and network conditions among surrogate nodes often lead to delays where slower nodes bottleneck the entire task. Existing systems lack a framework that accounts for both communication and computation costs when determining partition sizes.
This technology measures parameters for each surrogate node using training tasks and determines partition sizes based on a formula, ensuring all segment results return to the client simultaneously to minimize total task completion time. Applicable to augmented reality, real-time video analysis, and vehicular communication services using edge servers near base stations or Wi-Fi APs, it reduces both device battery consumption and response latency.
N/A
