Release Date: 14.10.2024
We are pleased to announce the release of Open Cluster Scheduler version 9.0.0, a pivotal evolution of widely respected projects, including Sun Grid Engine, Oracle Grid Engine, and Univa Grid Engine Open Core. This release introduces notable enhancements and new features designed to optimize workload management across diverse high-performance computing (HPC) and AI environments.
Key Features and Enhancements
- Advanced qconf Options: The introduction of new switches (ace, Ace, mce, Mce, sce, scel) enhances the management of resources, facilitating more precise and efficient configuration.
- Revamped Internal Architecture: The sge_qmaster now supports multiple data stores, laying the groundwork for improved scalability and future expansions.
- Resource Maps (RSMAP): Administrators can now define lists of specialized resources per host, such as GPUs and networking devices, through the new RSMAP resource type, enhancing resource allocation specificity.
- HOST Consumable Type: Introducing a new resource specification that allows resource allocation per host, independent of the number of (MPI) worker tasks, ideal for managing unique resources like GPUs and combined memory limits.
- JSON-Formatted Accounting Records: Transition to an efficient one-line JSONL format for accounting and reporting files simplifies data extensions, customizations, and facilitates smoother post-processing and exporting.
- Refined Parallel Job Resource Requests: Resource requests can now be defined at global, master, or slave levels, providing enhanced flexibility and control in parallel job management.
- High-Resolution Timestamps: With support for 64-bit timestamps, we address constraints like the year 2038 problem and facilitate finer-grained measurement capabilities.
- Enhanced qstat and qacct outputs to simplify dealing with complex workflows.
- Open-Source Innovations: Contributions such as open-sourced Java DRMAA, our test suite, and Go configuration APIs, underscore our commitment to community collaboration and transparency.
Expanded Platform Support
This release extends support to an array of platforms, including Linux-based AMD64, ARM, darwin-arm64, and FreeBSD, ensuring broader accessibility and integration flexibility.
Join Our Friendly Community
We believe that collaboration and mutual support are the keys to a successful open-source project. Your contributions and feedback are invaluable. Connect with us through:
- GitHub Repository: Access the source code, report issues, or contribute to the project. Open Cluster Scheduler on GitHub: https://github.com/hpc-gridware/clusterscheduler.
- Community Forums: Engage with other users and developers to exchange ideas and solutions on the open mailing list. Subscribe by sending an empty mail to: user+subscribe@clusterscheduler.org
Getting Started
Experience simplified HPC workload management:
- Download: Get Open Cluster Scheduler 9.0.0 from our website or build yourself from our open source repository.
- Install: Follow our step-by-step installation guide suited for your platform available in the documentation package or here.
- Explore: Dive into the new features and see how they can enhance your HPC environment.
Thank you for your continued support and commitment to advancing the capabilities of high-performance computing with Open Cluster Scheduler. We look forward to seeing the innovations and solutions you create with this new release.
Warm regards,
The HPC Gridware Team
