Serverless architecture enables various intelligent applications to be run without managing infrastructure. In this architecture, the computing cost is generally proportional to the number of requested stateless functions and this number can affect the task completion time, and thus it is prominent to decide an appropriate number of requested stateless functions. In this paper, we propose a latency-guaranteed and energy-efficient task offloading (LETO) system where an Internet of Things (IoT) device decides the number of stateless functions requested to the cloud by considering the deadline on the task completion time and its energy level. To minimize the computing cost while guaranteeing sufficiently short task completion time and low energy outage probability, we formulate a constrained Markov decision process (CMDP) problem and convert the CMDP problem into an equivalent linear programming (LP) model. By solving the LP model, the optimal policy on the number of requested stateless functions can be achieved. Evaluation results illustrate that LETO can cut down the operating expenditure (OPEX) by up to 59% compared to a latency-guaranteed offloading scheme while keeping the task completion time and the energy outage probability below desirable levels.
Bibliographical noteFunding Information:
This work was supported by the National Research Foundation (NRF) of Korea Grant funded by the Korean Government (MSIP) under Grant 2020R1A2C3006786 and Grant 2021R1A4A3022102.
- Computer architecture
- Energy harvesting
- Heuristic algorithms
- Serverless computing
- Task analysis
- Task offloading
- constrained Markov decision process (CMDP).
- energy harvesting
- serverless computing
ASJC Scopus subject areas
- Signal Processing
- Information Systems
- Hardware and Architecture
- Computer Science Applications
- Computer Networks and Communications