DREAMS: Decentralized Resource Allocation and Service Management across the Compute Continuum Using Service Affinity
Fuente:
arXiv
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916942382628864 |
|---|---|
| author | Dinh-Tuan, Hai Nguyen, Tien Hung Pandey, Sanjeet Raj |
| author_facet | Dinh-Tuan, Hai Nguyen, Tien Hung Pandey, Sanjeet Raj |
| contents | Modern manufacturing systems require adaptive computing infrastructures that can respond to highly dynamic workloads and increasingly customized production demands. The compute continuum emerges as a promising solution, enabling flexible deployment of microservices across distributed, heterogeneous domains. However, this paradigm also requires a novel approach to resource allocation and service placement, as traditional centralized solutions struggle to scale effectively, suffer from latency bottlenecks, and introduce single points of failure. In this paper, we present DREAMS, a decentralized framework that optimizes microservice placement decisions collaboratively across different computational domains. At its core, DREAMS introduces agents that operate autonomously within each domain while coordinating globally through a Raft-based consensus algorithm and cost-benefit voting. This decentralized architecture enables responsive, privacy-preserving, and fault-tolerant coordination, making it particularly suitable given the growing prevalence of multi-stakeholder scenarios across the compute continuum. In particular, within modern manufacturing environments, DREAMS achieves globally optimized service placements while maintaining high fault tolerance. Further evaluations demonstrate that key coordination operations, such as Local Domain Manager (LDM) registration and migration voting, scale sub-linearly with the number of domains, confirming the efficiency and scalability of our proposal. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_07497 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | DREAMS: Decentralized Resource Allocation and Service Management across the Compute Continuum Using Service Affinity Dinh-Tuan, Hai Nguyen, Tien Hung Pandey, Sanjeet Raj Distributed, Parallel, and Cluster Computing Performance Modern manufacturing systems require adaptive computing infrastructures that can respond to highly dynamic workloads and increasingly customized production demands. The compute continuum emerges as a promising solution, enabling flexible deployment of microservices across distributed, heterogeneous domains. However, this paradigm also requires a novel approach to resource allocation and service placement, as traditional centralized solutions struggle to scale effectively, suffer from latency bottlenecks, and introduce single points of failure. In this paper, we present DREAMS, a decentralized framework that optimizes microservice placement decisions collaboratively across different computational domains. At its core, DREAMS introduces agents that operate autonomously within each domain while coordinating globally through a Raft-based consensus algorithm and cost-benefit voting. This decentralized architecture enables responsive, privacy-preserving, and fault-tolerant coordination, making it particularly suitable given the growing prevalence of multi-stakeholder scenarios across the compute continuum. In particular, within modern manufacturing environments, DREAMS achieves globally optimized service placements while maintaining high fault tolerance. Further evaluations demonstrate that key coordination operations, such as Local Domain Manager (LDM) registration and migration voting, scale sub-linearly with the number of domains, confirming the efficiency and scalability of our proposal. |
| title | DREAMS: Decentralized Resource Allocation and Service Management across the Compute Continuum Using Service Affinity |
| topic | Distributed, Parallel, and Cluster Computing Performance |
| url | https://arxiv.org/abs/2509.07497 |