Semantic Translation of User Intent into Kubernetes Placement Constraints for AI Workflows

Abstract Modern AI applications are increasingly deployed across heterogeneous cloud–edge infrastructures, where each microservice in a workflow may require different resources and has different performance characteristics. Traditional container orchestration systems (e.g., Kubernetes) do not natively support either expressing high-level user goals or automatically translating them into concrete resource-aware deployment decisions. This paper presents CorMOS, a lightweight, intent-driven orchestration mechanism that enriches container deployment descriptors using ontology-based semantic knowledge. CorMOS introduces an OWL ontology for modelling workload types and user goals and applies semantic reasoning together with ML techniques to infer missing resource specifications and placement preferences. These inferred constraints are automatically injected into Helm charts and Argo Workflows, enabling Kubernetes to schedule microservices on the most appropriate cluster nodes. CorMOS was developed and evaluated both on a heterogeneous k3s testbed and within the NEMO Open Call 1 environment, where it was integrated into the NEMO orchestration pipeline. Experimental results show that CorMOS can autonomously complete incomplete workload descriptors, infer values for up to 50% of missing resource parameters while achieving 80% estimation accuracy, and incur low processing overhead while sustaining a throughput of approximately 20 charts/min. These findings demonstrate the effectiveness of semantic enrichment in reducing manual configuration effort while enabling intent-driven deployment with low computational overhead.

Authors

Publication Details

Journal
Research Square
Published
2026-11-16
DOI
https://doi.org/10.21203/rs.3.rs-11541809/v1
Primary Topic
Cloud Computing and Resource Management
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Semantic Translation of User Intent into Kubernetes Placement Constraints for AI Workflows

Evangelos Psomakelis, Efthymios Chondrogiannis, Efstathios Karanastasis
3 links
Research Square
Cloud Computing and Resource Management
preprint

Semantic Translation of User Intent into Kubernetes Placement Constraints for AI Workflows

Evangelos Psomakelis, Efthymios Chondrogiannis, Efstathios Karanastasis
preprint en

Abstract

Abstract Modern AI applications are increasingly deployed across heterogeneous cloud–edge infrastructures, where each microservice in a workflow may require different resources and has different performance characteristics. Traditional container orchestration systems (e.g., Kubernetes) do not natively support either expressing high-level user goals or automatically translating them into concrete resource-aware deployment decisions. This paper presents CorMOS, a lightweight, intent-driven orchestration mechanism that enriches container deployment descriptors using ontology-based semantic knowledge. CorMOS introduces an OWL ontology for modelling workload types and user goals and applies semantic reasoning together with ML techniques to infer missing resource specifications and placement preferences. These inferred constraints are automatically injected into Helm charts and Argo Workflows, enabling Kubernetes to schedule microservices on the most appropriate cluster nodes. CorMOS was developed and evaluated both on a heterogeneous k3s testbed and within the NEMO Open Call 1 environment, where it was integrated into the NEMO orchestration pipeline. Experimental results show that CorMOS can autonomously complete incomplete workload descriptors, infer values for up to 50% of missing resource parameters while achieving 80% estimation accuracy, and incur low processing overhead while sustaining a throughput of approximately 20 charts/min. These findings demonstrate the effectiveness of semantic enrichment in reducing manual configuration effort while enabling intent-driven deployment with low computational overhead.

Research Square
Cloud Computing and Resource Management
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon