Soft Actor-Critic-Based Service Migration in Multiuser MEC Systems DOI
Xinyu Zhang, Shuang Ren

Published: Dec. 22, 2023

Multi-access Edge Computing (MEC) is an emerging computing paradigm that provides abundant resource support for the next generation of Internet Things (IoT). When users move between edge servers with limited coverage, it necessary to determine service migration strategies ensure quality. However, traditional research on often oversimplifies multi-user scenario, focusing only individual users. This limitation hinders these methods from achieving optimality in real MEC environments. To address this issue, paper models scenario and focuses exploring impact user latency system energy consumption. We propose a method based discrete version Soft Actor-Critic algorithm (SACDM). Through simulation experiments evaluate performance our proposed solution, results demonstrate outperforms Deep Q-learning (DQNM) by reducing approximately 11.12%. Additionally, also achieves reduction consumption 6.66% compared DQNM.

Language: Английский

Editorial: Third Quarter 2023 IEEE Communications Surveys and Tutorials DOI Open Access
Dusit Niyato

IEEE Communications Surveys & Tutorials, Journal Year: 2023, Volume and Issue: 25(3), P. i - vi

Published: Jan. 1, 2023

I welcome you to the third issue of IEEE Communications Surveys and Tutorials in 2023. This includes 18 papers covering different aspects communication networks. In particular, these articles survey tutor various issues “Wireless Communications,” “Cyber Security,” “IoT M2M,” “Internet Technologies,” “Network Virtualization,” Service Management Green Communications.” A brief account for each is given below.

Language: Английский

Citations

0
Forecasting Trends in Cloud/Edge Computing: Unleashing the Power of Attention Mechanisms DOI Creative Commons
Berend Gort, María‐Antonia Serrano, Angelos Antonopoulos

et al.

Published: Nov. 7, 2023

The research in discussion explores the intersection of cloud/edge computing and time-series forecasting to optimize resource utilization reduce energy consumption telecommunications networks. It highlights evolution machine learning models used for forecasting, starting from Recurrent Neural Networks (RNNs) more advanced Long Short-Term Memory networks (LSTMs) with attention mechanisms, eventually transformer architectures. ultimate goal is achieve precise predictions allow smart cities telecom adapt real-time varying demands, improving service quality reducing operational costs. A significant focus given mechanism, especially sparse attention, which seen as a potential solution challenges faced by handling long sequences data efficiently. Among models, Informer model highlighted its promise domain scenarios. article also mentions providing list cutting-edge use cases proof-of-concept demonstrations substantiate claims regarding benefits these domain.

Language: Английский

Citations

0
Forecasting Trends in Cloud/Edge Computing: Unleashing the Power of Attention Mechanisms DOI Creative Commons
Berend Gort, María‐Antonia Serrano, Angelos Antonopoulos

et al.

Published: Nov. 7, 2023

<p>The research in discussion explores the intersection of cloud/edge computing and time-series forecasting to optimize resource utilization reduce energy consumption telecommunications networks. It highlights evolution machine learning models used for forecasting, starting from Recurrent Neural Networks (RNNs) more advanced Long Short-Term Memory networks (LSTMs) with attention mechanisms, eventually transformer architectures. The ultimate goal is achieve precise predictions allow smart cities telecom adapt real-time varying demands, improving service quality reducing operational costs.</p> <p>A significant focus given mechanism, especially sparse attention, which seen as a potential solution challenges faced by handling long sequences data efficiently. Among models, Informer model highlighted its promise domain scenarios. article also mentions providing list cutting-edge use cases proof-of-concept demonstrations substantiate claims regarding benefits these domain.</p>

Language: Английский

Citations

0
Distributed Intelligence for Dynamic Task Migration in the 6G User Plane using Deep Reinforcement Learning DOI Creative Commons
Sayantini Majumdar, Susanna Schwarzmann, Riccardo Trivisonno

et al.

Published: Nov. 7, 2023

<p>In-Network Computing (INC) is a currently emerging paradigm. Realizing INC in 6G networks could mean that user plane entities (UPEs) carry out computations on packets while transmitting them. These may have specific requirements terms of their completion time. In case high compute pressure at one UPE, migrating to another UPE be beneficial, order avoid exceeding the time requirement. Centralized migration approaches suffer from increased signaling and are prone react too slow. Therefore, this paper investigates applicability distributed intelligence tackle problem task plane. Each equipped with an intelligent agent, enabling autonomous decisions whether should migrated UPE. To enable agents learn apply optimal policy, we investigate compare two state-of-the-art Deep Reinforcement Learning (DRL) approaches: Advantage Actor-Critic (A2C) Double Q-Network (DDQN). We show, via simulations, performance both solutions, percentage tasks requirement, near-optimal training A2C least 60% faster. </p>

Language: Английский

Citations

0
Distributed Intelligence for Dynamic Task Migration in the 6G User Plane using Deep Reinforcement Learning DOI Creative Commons
Sayantini Majumdar, Susanna Schwarzmann, Riccardo Trivisonno

et al.

Published: Nov. 7, 2023

<p>In-Network Computing (INC) is a currently emerging paradigm. Realizing INC in 6G networks could mean that user plane entities (UPEs) carry out computations on packets while transmitting them. These may have specific requirements terms of their completion time. In case high compute pressure at one UPE, migrating to another UPE be beneficial, order avoid exceeding the time requirement. Centralized migration approaches suffer from increased signaling and are prone react too slow. Therefore, this paper investigates applicability distributed intelligence tackle problem task plane. Each equipped with an intelligent agent, enabling autonomous decisions whether should migrated UPE. To enable agents learn apply optimal policy, we investigate compare two state-of-the-art Deep Reinforcement Learning (DRL) approaches: Advantage Actor-Critic (A2C) Double Q-Network (DDQN). We show, via simulations, performance both solutions, percentage tasks requirement, near-optimal training A2C least 60% faster. </p>

Language: Английский

Citations

0
Soft Actor-Critic-Based Service Migration in Multiuser MEC Systems DOI
Xinyu Zhang, Shuang Ren

Published: Dec. 22, 2023

Multi-access Edge Computing (MEC) is an emerging computing paradigm that provides abundant resource support for the next generation of Internet Things (IoT). When users move between edge servers with limited coverage, it necessary to determine service migration strategies ensure quality. However, traditional research on often oversimplifies multi-user scenario, focusing only individual users. This limitation hinders these methods from achieving optimality in real MEC environments. To address this issue, paper models scenario and focuses exploring impact user latency system energy consumption. We propose a method based discrete version Soft Actor-Critic algorithm (SACDM). Through simulation experiments evaluate performance our proposed solution, results demonstrate outperforms Deep Q-learning (DQNM) by reducing approximately 11.12%. Additionally, also achieves reduction consumption 6.66% compared DQNM.

Language: Английский

Citations

0