Task Scheduling Dataset for Cloud Environments

A collection of workflow task graphs for benchmarking and evaluating task scheduling algorithms in heterogeneous computing environments.

Dataset Structure

Each top-level directory represents a scientific workflow application. Subdirectories within each workflow are organized by the number of nodes (tasks) in the workflow graph.

Workflow	Description	Node Sizes Available
FFT	Fast Fourier Transform workflows	40, 96, 224, 512, 1152
GE	Gene Expression analysis workflows	9, 35, 135, 527, 2079
Genome	Genome sequencing workflows	50 – 1000 (in increments of 50–100)
LA	Linear Algebra workflows	11, 37, 137, 529, 2081
LIGO	Laser Interferometer Gravitational-Wave Observatory workflows	50 – 1000 (in increments of 50–100)
Montage	Astronomical image mosaicking workflows	50 – 1000 (in increments of 50–100)

Directory Layout

Dataset/
├── FFT/
│   ├── 40Nodes/
│   ├── 96Nodes/
│   ├── 224Nodes/
│   ├── 512Nodes/
│   └── 1152Nodes/
├── GE/
│   ├── 9Nodes/
│   ├── 35Nodes/
│   ├── 135Nodes/
│   ├── 527Nodes/
│   └── 2079Nodes/
├── Genome/
│   ├── 50Nodes/
│   ├── 100Nodes/
│   ├── ...
│   └── 1000Nodes/
├── LA/
│   ├── 11Nodes/
│   ├── 37Nodes/
│   ├── 137Nodes/
│   ├── 529Nodes/
│   └── 2081Nodes/
├── LIGO/
│   ├── 50Nodes/
│   ├── 100Nodes/
│   ├── ...
│   └── 1000Nodes/
└── Montage/
    ├── 50Nodes/
    ├── 100Nodes/
    ├── ...
    └── 1000Nodes/

Usage

These datasets can be used to evaluate scheduling algorithms on metrics such as:

• Makespan — total execution time of the workflow
• Resource utilization — CPU/memory efficiency across cloud instances
• Cost — monetary cost of cloud resource allocation
• Scalability — algorithm performance as workflow size increases

The varying node counts within each workflow allow for scalability testing — from small workflows (9–40 nodes) to large-scale workflows (1000–2081 nodes).