WORKSWORLD: A Domain for Integrated Numeric Planning and Scheduling of Distributed Pipelined Workflows

Taylor Paul; William Regli

cs.DCMar 12, 2026

WORKSWORLD: A Domain for Integrated Numeric Planning and Scheduling of Distributed Pipelined Workflows

Authors: Taylor Paul, William Regli

Organizations: University of Maryland, College Park, MD USA

Abstract

This work pursues automated planning and scheduling of distributed data pipelines, or workflows. We develop a general workflow and resource graph representation that includes both data processing and sharing components with corresponding network interfaces for scheduling. Leveraging these graphs, we introduce WORKSWORLD, a new domain for numeric domain-independent planners designed for permanently scheduled workflows, like ingest pipelines. Our framework permits users to define data sources, available workflow components, and desired data destinations and formats without explicitly declaring the entire workflow graph as a goal. The planner solves a joint planning and scheduling problem, producing a plan that both builds the workflow graph and schedules its components on the resource graph. We empirically show that a state-of-the-art numeric planner running on commodity hardware with one hour of CPU time and 30GB of memory can solve linear-chain workflows of up to 14 components across eight sites.

Figures & tables

Panel of Figure 1 — Figure 1: The Worksworld framework. Primary research contributions are green; secondary engineering artifacts provided without experimental evaluation are yellow.

Figure 2: Our scheduling problem consists of mapping movable components in the left-hand DAG to utilize compute, storage and network resources across sites.

Figure 3: Sites and the relevant annotations and resources for our scheduling problem.

Panel of Figure 4 — Figure 4: PDDL type hierarchy for Worksworld and abbreviations utilized throughout paper.

Predicates ( $\mathcal{F}$ )	Numeric Fluents ( $\mathcal{X}$ )
(available_at I R S)	(resource_total I R)
(scheduled_on WC IF)	(resource_available I R)
(type_of WC WCT)	(work_amount WC R)
(fixed WC)	(msg_max_rate WC)
(linked L S S)	(msg_actual_rate WC IF)
(link_uses CL DL)	(msg_size DCT)

Table 1: Available predicates and numeric fluents in Worksworld .

Actions ( $\mathcal{A}$ ) & Input Object Types
(replicate_code WC S S L)
(schedule_component WC R IF R S)
(connect_direct_link
PC DPI S DC DCT DSI S DL DIR)
(connect_composite_link
PC DPI S DC DCT DSI S CL DL DL DIR)

Table 2: Available actions in Worksworld to build and schedule pipelined workflows.

Object Type	F	X	A
Site	0	0	$\mathcal{O}(n^{2})$
Direct Link	2	3	$\mathcal{O}(n^{2})$
Composite Link	4	3	$\mathcal{O}(n)$
Interface ( DSI / DPI )	1	2	$\mathcal{O}(n^{2})$
Data Component	2	3	$\mathcal{O}(n^{2})$
Processing Component	4	3	$\mathcal{O}(n^{2})$

Table 3: Additional predicates (F), numeric fluents (X), and grounded actions (A) required by additional object types.

Problem Size	F (AIBR)	X (AIBR)	A (AIBR)
2WFC, 1S	22 (8)	7 (6)	14 (6)
2WFC, 2S	96 (24)	14 (12)	90 (22)
2WFC, 3S	306 (39)	22 (19)	284 (51)
4WFC , 3S	843 (81)	28 (25)	1146 (109)

Table 4: Grounding comparison: Predicates (F), Numeric Fluents (X), and Grounded Actions (A) before and after AIBR preprocessing.

Figure 5: Planning and scheduling linear-chain workflows varying one aspect: workflow components, interfaces, direct links and sites.

Figure 6: Planning and scheduling variable length linear-chain workflows on a complex resource graph (8S, 16IF, 15DL, 14CL).

Workflow Automation

Workflow automation coordinates multi-step tasks across applications, tools, and people using software, AI agents, and explicit process logic.

See all 62 papers

Cloud Workflow Scheduling Based on Graph Attention-Driven Hierarchical Reinforcement Learning
Sep 14, 2026Zongjin Li, Shaohan Feng, Chunxi Yang +1Real-Time Scheduling Hierarchical RL
Faculty of Mechanical and Electrical Engineering, Kunming University of Science and Technology, Kunming 650500, China · School of Information and Electronic Engineering (Sussex Artificial Intelligence Institute), Zhejiang Gongshang University, Hangzhou 310018, China
Planning and Scheduling Business Processes under Control-Flow Uncertainty: Extended Version
Sep 4, 2026Michel Kunkler, Stefanie Rinderle-MaStochastic Optimization Chance-Constrained Optimization
Technical University of Munich, Germany TUM School of Computation, Information, and Technology
Duration-Informed Workload Scheduler
Mar 7, 2026Daniela Loreti, Davide Leone, Andrea BorghesiHigh-Performance Computing
DISI, University of Bologna, Bologna, Viale Risorgimento 2, Italy

Sep 14, 2026J/K move · Enter open · S save

Cloud Workflow Scheduling Based on Graph Attention-Driven Hierarchical Reinforcement Learning

Zongjin Li, Shaohan Feng, Chunxi Yang, Wenbo Wang

Faculty of Mechanical and Electrical Engineering, Kunming University of Science and Technology, Kunming 650500, China · School of Information and Electronic Engineering (Sussex Artificial Intelligence Institute), Zhejiang Gongshang University, Hangzhou 310018, China

Real-Time Scheduling Hierarchical RL

Open paper AskPDF arXiv

Dynamic cloud workflow scheduling must balance deadline satisfaction, container utilization, and energy consumption while dealing with stochastic task-execution speeds, placement-dependent communication, and coupled task and container decisions. Workflows are naturally modeled as directed acyclic graphs (DAGs), but conventional vector- or matrix-based states do not fully capture their dependency topology. To better represent task urgency and structural relationships, we assign predicted sub-deadlines to tasks and use a multi-head graph attention network (GAT) to extract dependency information from the evolving DAGs. Based on these representations, we develop a Graph Attention-Driven Hierarchical Reinforcement Learning (GA-HRL) framework and model the scheduling process as an event-driven hierarchical semi-Markov decision process (SMDP). Workflow arrivals and task completions trigger scheduling events. At each scheduling event, the Task Scheduling (TS) agent first processes the currently ready tasks by assigning them to admissible existing containers or requesting new ones. The requested containers are then processed by the Container Scheduling (CS) agent for host placement before the environment advances. The two agents are trained alternately using separate Proximal Policy Optimization (PPO). Experiments on the 2018 Alibaba cluster trace show that GA-HRL maintains competitive workflow success rate and, in settings where success is comparable, generally achieves higher container utilization and lower energy consumption. Under the largest speed variation, it trades a small success-rate margin for substantially lower energy. Simulation code is available at: https://github.com/zongjin130/GA-HRL.

Abstract

Workflow Automation

Explore similar work

Cloud Workflow Scheduling Based on Graph Attention-Driven Hierarchical Reinforcement Learning

Planning and Scheduling Business Processes under Control-Flow Uncertainty: Extended Version

Duration-Informed Workload Scheduler