A density-aware and overlap-driven framework for task allocation in distributed spatial data processing environments

doi:10.1016/j.future.2026.108861

A density-aware and overlap-driven framework for task allocation in distributed spatial data processing environments

2026-10-09

0

OA

AI

Christoforos Kosmıdis

K

Konstantinos Kolomvatsos

DOI：10.1016/j.future.2026.108861

Original

Original request for help

Save

Abstract

En 中文

The increasing volume of data generated by distributed sensing and Internet of Things (IoT) devices and applications has intensified the need for efficient data and task management in edge and distributed environments. A key challenge in such systems is determining how to allocate tasks that require the processing of multidimensional data across distributed nodes while preserving data locality and limiting communication overhead. In this paper, we propose a density-aware and overlap-driven framework for task allocation in distributed data-processing environments. Tasks and node data regions are represented as multidimensional hyper-rectangles, while each node maintains information about its local data density. The proposed framework employs a scenario-based allocation mechanism based on strict spatial containment, tolerance-based containment, and external execution. Depending on the spatial relationship between a task and the available data regions, as well as the corresponding density conditions, the task can be executed locally, supported through neighboring-node collaboration, offloaded to another node, or reported as unserved when no suitable execution node is available. Candidate nodes for collaboration and offloading are selected using a scoring mechanism that considers spatial overlap and communication latency. To improve adaptability under heterogeneous data distributions, the framework additionally supports adaptive density thresholds and bounded tolerance expansion according to the observed node densities. The proposed approach is evaluated using three heterogeneous real-world datasets: OpenStreetMap (OSM) data from Greece, NYC Yellow Taxi data, and data derived from the National Ambulatory Medical Care Survey (NAMCS). The evaluation examines allocation behavior, execution success, communication requirements, decision overhead, scalability, parameter sensitivity, and deadline-aware admission. The results demonstrate accurate allocation decisions with low decision overhead while highlighting the trade-offs among local execution, neighbor collaboration, and external offloading. The proposed framework provides an interpretable and adaptive approach to multidimensional task allocation and offers a practical basis for distributed data processing in edge and IoT environments.

Keywords:

Edge computing

Distributed data processing

Task allocation

Spatial data management

Data density

Task offloading

Internet of things

A density-aware and overlap-driven framework for task allocation in distributed spatial data processing environments

Abstract

Journal

Organization

Cited Papers