SAGAZ
Interactive explorer for SAGA — Scheduling Algorithms Gathered. Browse algorithms, datasets, benchmarking results, and adversarial analysis.
Explore SAGA
Algorithm Catalog
Browse 22+ scheduling algorithms with interactive examples showing how each one schedules tasks across processors.
Dataset Explorer
Explore 16 diverse datasets from random graphs and scientific workflows to IoT streaming applications.
Benchmarking Results
Interactive heatmaps showing how algorithms compare across datasets with drill-down capability.
PISA Analysis
Adversarial analysis revealing where algorithms fail. For every algorithm, PISA finds instances where it performs 2x+ worse.
What is SAGA?
SAGA (Scheduling Algorithms Gathered) is a Python toolkit for designing, comparing, and visualizing DAG-based computational workflow scheduler performance on heterogeneous compute networks.
The problem: given a set of interdependent tasks (a Directed Acyclic Graph) and a heterogeneous network of processors with varying speeds and communication bandwidths, find a schedule that minimizes the total execution time (makespan).
SAGA provides a unified API for 22+ scheduling algorithms, 16 diverse datasets, and the PISA (Problem Instance Simulated Annealing) framework for adversarial analysis — finding problem instances where algorithms perform surprisingly poorly.