Abstract
With the wide deployment of computing capabilities and artificial intelligence (AI) techniques, Internet of intelligence is envisioned as a natural tendency in future networks. This technological revolution will foster abundant computing services with new requirements, such as distributed large AI model training and on-the-fly image rendering, which require dynamic collaboration of multi-dimensional resources from both the communication and computation perspectives. In this article, we introduce a communication and computation integrated network architecture, named (Com)2Net. Enormous computing traffic can be scheduled across space-air-ground domain, end-edge-cloud domain, and multi-data center domain, which facilitates ubiquitous connectivity and collaborative computation, thus supporting diverse advanced computing services. Furthermore, an intelligent resource adaption scheme is proposed to dynamically orchestrate multi-dimensional resources for massive concurrent computing tasks, in which quantum genetic algorithms are utilized to make the best joint decision for inter-domain traffic scheduling and intra-domain resource allocation. Finally, we present a case study, and discuss open research issues that are fundamental for efficient collaborative computation in (Com)2Net.
Original language | English |
---|---|
Pages (from-to) | 1 |
Number of pages | 1 |
Journal | IEEE Network |
DOIs | |
Publication status | Accepted/In press - 2024 |
Keywords
- Collaboration
- Computer architecture
- Processor scheduling
- Quality of service
- Servers
- Task analysis
- Training