Code of Paper: VoI-Guaranteed Task Computing for Massive IoT under Demand and Resource Uncertainties

Abstract—In massive Internet of Things (IoT) deployments,

the efficient allocation of computing resources to IoT devices

while preserving devices’ data poses a significant challenge.

This paper proposes a new online probabilistic model to address

uncertainties in demand and resource allocation for IoT

networks, where the task computing of requesting devices is

addressed by serving devices. The model incorporates uncertainty

and formulates an optimization problem, concerning available

computing resources, aimed at minimizing the number of devices

used for task offloading. In addition, the Value of Information

(VoI) is considered to ensure that the utility of information

contained in the data associated with each device remains above

a certain threshold. The problem is solved using a heuristic

algorithm, where if there is no server available to respond,

the tasks are stored in a buffer for the next time slot, and

their waiting time is considered. This approach is based on bin

packing algorithms, which have demonstrated their effectiveness

in optimizing resource utilization and task allocation. Furthermore,

feasibility analyses are provided to assess the performance

of the optimal algorithm in the worst case scenarios, thereby

ensuring the best possible outcomes. In addition, we investigate

two strategies for addressing the buffered tasks, First in First Out

(FIFO) and Last in First Out (LIFO), to model the task response,

where the adoption of FIFO reduces the average waiting time by

a factor of two. In general, the proposed framework delivers a

high level of task computing service reliability, with each serving

device accommodating an average of four requesting devices.

Index Terms— Online task offloading, Distributed resource allocation,

Demand uncertainty, Resource uncertainty, and Resource

management.