Cloud computing has emerged as a vital platform for processing resource-intensive workloads in smart manu-facturing environments,enabling scalable and flexible access to remote data centers over the internet.In these ...Cloud computing has emerged as a vital platform for processing resource-intensive workloads in smart manu-facturing environments,enabling scalable and flexible access to remote data centers over the internet.In these environments,Virtual Machines(VMs)are employed to manage workloads,with their optimal placement on Physical Machines(PMs)being crucial for maximizing resource utilization.However,achieving high resource utilization in cloud data centers remains a challenge due to multiple conflicting objectives,particularly in scenarios involving inter-VM communication dependencies,which are common in smart manufacturing applications.This manuscript presents an AI-driven approach utilizing a modified Multi-Objective Particle Swarm Optimization(MOPSO)algorithm,enhanced with improved mutation and crossover operators,to efficiently place VMs.This approach aims to minimize the impact on networking devices during inter-VM communication while enhancing resource utilization.The proposed algorithm is benchmarked against other multi-objective algorithms,such as Multi-Objective Evolutionary Algorithm with Decomposition(MOEA/D),demonstrating its superiority in optimizing resource allocation in cloud-based environments for smart manufacturing.展开更多
In order to achieve global multiple seamless coverage,space-based internet usually adopts low Earth orbit(LEO)mega-constellation networks structure,which has the characteristics of high network topology dynamics,limit...In order to achieve global multiple seamless coverage,space-based internet usually adopts low Earth orbit(LEO)mega-constellation networks structure,which has the characteristics of high network topology dynamics,limited on-board computing and storage capacity,and uneven distribution of ground traffic.Such features may cause problems such as high transmission delay,network congestion and link interruption.Establishing a stable,efficient and balanced satellite communication link can effectively alleviate the performance of the transmission delay,load balancing,and network throughput.Taking advantage of the regularity of network topology,a pre-coded inter-satellite routing algorithm with load balancing is proposed,which includes 3 parts:(a)the routing sequence coding method and the concept of gateway satellite Service Region(GSSR)are proposed;(b)the initial routing sequence of GSSR is generated based on the maximum network flow method under the ideal situation of uniform satellite traffic distribution;(c)aiming at the uneven distribution of traffic,the Sinkhorn algorithm is used to improve the load balancing performance of inter-satellite links.Simulation results show that,for the Starlink Group-4 constellation,the proposed method can maintain a low transmission delay and improve the load balancing together with the network throughout performance with minimal hops and low time complexity.展开更多
基金funded by Researchers Supporting Project Number(RSPD2025R 947),King Saud University,Riyadh,Saudi Arabia.
文摘Cloud computing has emerged as a vital platform for processing resource-intensive workloads in smart manu-facturing environments,enabling scalable and flexible access to remote data centers over the internet.In these environments,Virtual Machines(VMs)are employed to manage workloads,with their optimal placement on Physical Machines(PMs)being crucial for maximizing resource utilization.However,achieving high resource utilization in cloud data centers remains a challenge due to multiple conflicting objectives,particularly in scenarios involving inter-VM communication dependencies,which are common in smart manufacturing applications.This manuscript presents an AI-driven approach utilizing a modified Multi-Objective Particle Swarm Optimization(MOPSO)algorithm,enhanced with improved mutation and crossover operators,to efficiently place VMs.This approach aims to minimize the impact on networking devices during inter-VM communication while enhancing resource utilization.The proposed algorithm is benchmarked against other multi-objective algorithms,such as Multi-Objective Evolutionary Algorithm with Decomposition(MOEA/D),demonstrating its superiority in optimizing resource allocation in cloud-based environments for smart manufacturing.
基金supported by the Key Research and Development Program of Hubei Province(grant number 2022BID017)the Key-Area Research and Development Program of Guangdong Province(grant numbers 2019B010158001)the Scientific Research Project of National University of Defense Technology(No.ZK22-02).
文摘In order to achieve global multiple seamless coverage,space-based internet usually adopts low Earth orbit(LEO)mega-constellation networks structure,which has the characteristics of high network topology dynamics,limited on-board computing and storage capacity,and uneven distribution of ground traffic.Such features may cause problems such as high transmission delay,network congestion and link interruption.Establishing a stable,efficient and balanced satellite communication link can effectively alleviate the performance of the transmission delay,load balancing,and network throughput.Taking advantage of the regularity of network topology,a pre-coded inter-satellite routing algorithm with load balancing is proposed,which includes 3 parts:(a)the routing sequence coding method and the concept of gateway satellite Service Region(GSSR)are proposed;(b)the initial routing sequence of GSSR is generated based on the maximum network flow method under the ideal situation of uniform satellite traffic distribution;(c)aiming at the uneven distribution of traffic,the Sinkhorn algorithm is used to improve the load balancing performance of inter-satellite links.Simulation results show that,for the Starlink Group-4 constellation,the proposed method can maintain a low transmission delay and improve the load balancing together with the network throughout performance with minimal hops and low time complexity.
