# A Learning Method with Gap-Aware Generation for Heterogeneous DAG Scheduling Source: [A Learning Method with Gap-Aware Generation for Heterogeneous DAG Scheduling]() Collection index: [llms.txt]() Data updated: 2026-06-17T05:00:05.807Z Authors: Ruisong Zhou, Haijun Zou, Li Zhou, Chumin Sun, Zaiwen Wen Published: 2026-03-24T14:16:08.000Z ## Abstract Efficient scheduling of directed acyclic graphs (DAGs) is a core problem in large-scale data-intensive computing systems, where query plans, data-processing workloads, and computation graphs consist of dependent tasks competing for limited heterogeneous resource pools. In practice, achieving high-performance execution requires schedulers to adapt across environments with varying resource pools and task types, while generating schedules under tight runtime budgets. We propose WeCAN, an end-to-end reinforcement learning framework for heterogeneous DAG scheduling that addresses task-pool compatibility coefficients and generation-induced optimality gaps. It adopts a two-stage single-pass design: a single forward pass produces task-pool scores and global parameters, followed by a generation map that constructs schedules without repeated network calls. Its weighted cross-attention encoder models task-pool interactions gated by compatibility coefficients, and is size-agnostic to environment fluctuations. Moreover, widely used list-scheduling maps can incur generation-induced optimality gaps from restricted reachability. We introduce an order-space analysis that characterizes the reachable set of generation maps via feasible schedule orders, explains the mechanism behind generation-induced gaps, and yields sufficient conditions for gap elimination. Guided by these conditions, we design a skip-extended realization with an analytically parameterized decreasing skip rule, which enlarges the reachable order set while preserving single-pass efficiency. Experiments on real-world TPC-H query DAGs, resource-intensive workload datasets, and ML-compiler computation graphs demonstrate improved makespan over strong baselines, with inference time comparable to classical heuristics and faster than multi-round neural schedulers. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [DAG Scheduling on TPC-H Full Generalization 150 tasks](): DAG scheduling; TPC-H Full Generalization 150 tasks; Makespan: 24338. - [DAG Scheduling on TPC-H Full Generalization 200 tasks](): DAG scheduling; TPC-H Full Generalization 200 tasks; Makespan: 31422. - [DAG Scheduling on Erdős-Rényi 500 tasks](): DAG scheduling; Erdős-Rényi 500 tasks; Makespan: 10083. - [DAG Scheduling on Layer Graphs 500 tasks](): DAG scheduling; Layer Graphs 500 tasks; Makespan: 10752. - [DAG Scheduling on Stochastic Block 500 tasks](): DAG scheduling; Stochastic Block 500 tasks; Makespan: 10019. - [Heterogeneous Resource Scheduling on TPC-H 30 (3 pools)](): Heterogeneous Resource Scheduling; TPC-H 30 (3 pools); Makespan: 18964. - [Heterogeneous Resource Scheduling on TPC-H 50 (3 pools)](): Heterogeneous Resource Scheduling; TPC-H 50 (3 pools); Makespan: 32814. - [Heterogeneous Resource Scheduling on TPC-H-100 (3 pools)](): Heterogeneous Resource Scheduling; TPC-H-100 (3 pools); Makespan: 61373. - [Computation graph scheduling on TPC-H heavy 30](): Computation graph scheduling; TPC-H heavy 30; Makespan: 24099. - [Computation graph scheduling on TPC-H 50 heavy](): Computation graph scheduling; TPC-H 50 heavy; Makespan: 35695. ## Original sources and implementation [Original paper]() [PDF]()