|
Work item:
|
Y.fAIAC-reqts
|
|
Subject/title:
|
Cloud computing - Functional requirements for federation of distributed AI accelerators
|
|
Status:
|
Under study
|
|
Approval process:
|
AAP
|
|
Type of work item:
|
Recommendation
|
|
Version:
|
New
|
|
Equivalent number:
|
-
|
|
Timing:
|
2028-Q4 (High priority)
|
|
Liaison:
|
ITU-T SG5, SG11, SG17, SG21, ISO/IEC JTC 1/SC 38, SC 42
|
|
Supporting members:
|
Kyung Hee University, ETRI, Korea (Republic of)
|
|
Summary:
|
As AI applications expand across cloud environments, the demand for AI services is growing, leading cloud service providers (CSPs) to support large-scale AI training and inference services. However, AI services based on large-scale models, such as large language models (LLMs), require substantial AI accelerators. To address these requirements, computing resources deployed in datacentre, cloud, and on-premises are increasingly utilized to expand the available resource beyond a single domain. Despite these advancements, challenges remain in utilizing distributed AI accelerator resources efficiently, including heterogeneous AI accelerator characteristics, uneven resource distribution, separated resource management, and resource fragmentation across different domains. Effective AI accelerator federation to provide various federated resource volumes to satisfy CSCs objectives is crucial to improving cost-effectiveness, performance, reliability, and resource utilization. Therefore, we propose developing a draft Recommendation on functional requirements for federation of distributed AI accelerators.
|
|
Comment:
|
-
|
|
Reference(s):
|
|
|
Historic references:
|
|
Contact(s):
|
|
| ITU-T A.5 justification(s): |
|
|
|
|
First registration in the WP:
2026-07-23 14:24:26
|
|
Last update:
2026-07-23 14:29:44
|
|
