A new scheduling system named the context-aware timeline scheduler predicts computational resource demands and execution makespan for data pipelines operating across edge, fog, and cloud infrastructures, according to a study published in the Journal of Cloud Computing. Tested on an industrial spot welding data preparation pipeline intended for downstream artificial intelligence and machine learning analytics, the framework demonstrated accurate forecasting of resource consumption and total makespan for target data volumes up to two orders of magnitude higher than the low-scale inputs measured during preliminary testing.
The increasing adoption of the Internet of Things is driving expanded reliance on data pipelines to process distributed data across the computing continuum. In modern computing environments, these data pipelines are structured as directed acyclic graphs consisting of heterogeneous processing steps that transform raw data streams to generate business value. However, practical production pipelines cannot undergo comprehensive benchmarking in advance, creating a technical bottleneck that prevents system operators from deploying traditional scheduling algorithms.
To overcome the challenge of absent benchmark data, researchers developed the context-aware timeline scheduler, designated as CATS. During these dry-run procedures, consecutive test deployments of individual pipeline steps run on the designated target infrastructure, providing empirical measurements of central processing unit activity, memory usage patterns, and individual step execution times.
Using the empirical observations collected from dry-run testing, CATS incorporates pipeline structure and data locality information to anticipate operational behavior at full production scale. The system evaluates the captured metrics to predict resource requirements, execution time, and the necessary degree of horizontal scaling for each processing step when managing larger total input volumes. Through this predictive mechanism, the scheduler aims to optimize overall makespan and deployment expenditure across distributed continuum resources.
After establishing step-level predictions for execution time and scaling requirements, the scheduler generates a concrete timeline of scheduling events. This timeline defines the orchestration sequence executed when deploying the data pipeline across target infrastructure components.
The authors evaluated the scheduling framework on physical infrastructure configured to emulate an integrated fog and cloud computing continuum. The evaluation utilized an industrial spot welding data preparation pipeline configured for downstream predictive analytics using machine learning models. Experimental results confirmed high prediction accuracy for central processing unit and memory usage as well as total execution makespan when scaling data inputs up to two orders of magnitude beyond the dry-run baseline.
The team noted that the dry-run methodology enables effective scheduling across edge, fog, and cloud continuum resources without requiring prior full-scale benchmarking.
Subject of Research: Data pipeline scheduling across edge, fog, and cloud computing continuums
Article Title: CATS: context-aware timeline scheduler for data pipelines on the computing continuum
Article References: Nikolov, N., Mehran, N., Soylu, A., Prodan, R., & Roman, D. (2026). CATS: context-aware timeline scheduler for data pipelines on the computing continuum. Journal of Cloud Computing. https://doi.org/10.1186/s13677-026-01019-9
Image Credits: AI Generated
DOI: 10.1186/s13677-026-01019-9
Keywords: data pipelines, computing continuum, timeline scheduler, cloud computing, edge computing, fog computing, resource prediction, makespan optimization, directed acyclic graphs, dry runs, data locality, horizontal scaling, industrial spot welding, predictive analytics, machine learning, Internet of Things
News Source: Denise Maddox. (November 17, 2026). Context-Aware Timeline Scheduler Accurately Predicts Data Pipeline Performance on Computing Continuums. Scienmag.



