Moab Workload Manager

Moab Workload Manager
Name	Moab Workload Manager
Developer	Adaptive Computing
Released	0 2001
Latest release version	10.0.0
Latest release date	October 2022
Operating system	Linux, Unix
Genre	Job scheduler, Workload manager
License	Proprietary software
Website	https://www.adaptivecomputing.com

Moab Workload Manager. It is a comprehensive job scheduler and workload manager designed for high-performance computing (HPC) environments, cluster computing, and grid computing. Originally developed by the Cluster Resources company, which later became Adaptive Computing, it provides centralized management for complex computing resources and scientific workloads. The software is known for its sophisticated policy engine and deep integration with other HPC middleware and infrastructure software.

Overview

Moab Workload Manager functions as an intelligent meta-scheduler, coordinating jobs across diverse resources like those found in a supercomputer or a data center. It evolved from earlier batch scheduler technologies to address the growing complexity of scientific computing and enterprise technical computing. The system interacts with local resource managers such as TORQUE Resource Manager (originally an open-source spinoff from Moab), Slurm Workload Manager, and IBM Platform LSF to execute and monitor jobs. Its development has been closely tied to projects funded by agencies like the United States Department of Energy and the National Science Foundation.

Architecture and Components

The core architecture is modular, centering on the Moab server daemon which makes scheduling decisions. It communicates with resource managers via plugins and uses a dedicated MySQL or PostgreSQL database for storing job metadata and configuration. Key internal components include the Policy Engine, which evaluates complex rules, and the Analytics Module for reporting. For grid computing, it can integrate with the Globus Toolkit, and for cloud computing, it interfaces with platforms like OpenStack and Amazon Web Services. The Moab Accounting Manager (MAM) handles detailed chargeback and reporting.

Features and Capabilities

Its advanced features include backfill scheduling to improve system utilization, fairshare scheduling to enforce organizational policies, and preemptive scheduling for high-priority work. The software supports advanced reservations for guaranteed resource access, critical for time-sensitive projects. It provides robust quality of service (QoS) management, power management capabilities for green computing, and topology-aware scheduling to optimize for NUMA architectures or GPU placements. Workload profiling and predictive analytics help in capacity planning.

Configuration and Management

Administration is performed through configuration files, a command-line interface, and a web-based portal known as the Moab Web Suite. Policies are defined using a specialized policy language to control job priority, fairshare, and access control lists. The system allows for the creation of virtual private clusters within a larger HPC cluster, enabling multi-tenant environments for different departments or projects like those at Lawrence Livermore National Laboratory or Purdue University.

Integration and Interfaces

Moab is designed for broad interoperability within the HPC ecosystem. Its native integration with TORQUE Resource Manager is well-known, but it also provides connectors for Altair PBS Professional, Slurm Workload Manager, and Microsoft HPC Pack. For grid and multi-cluster environments, it supports standards from the Open Grid Forum. Monitoring integrations exist with tools like Ganglia and Nagios, and its REST API allows for custom integration with external orchestration software and service portals.

Use Cases and Applications

Primary deployments are in large-scale scientific research facilities, such as national laboratories like Oak Ridge National Laboratory and Argonne National Laboratory, and major academic research centers supporting projects in computational fluid dynamics, genomics, and climate modeling. It is also used in computer-aided engineering (CAE) for simulation workloads in industries like aerospace and automotive manufacturing. Furthermore, it manages rendering farms for visual effects companies and supports financial modeling in investment banking.

Category:Job scheduling software Category:High-performance computing Category:Linux software Category:Proprietary software