Gridware Cluster Scheduler

Gridware Cluster Scheduler is the official successor of what has been known as "Grid Engine" (SGE, UGE, etc.). It is developed by the original developers of Sun Grid Engine and Univa Grid Engine. Gridware Cluster Scheduler provides 100% compatible "SGE" interfaces while improving scalability, platform support, and offering unique and modern features. It is fully supported by HPC Gridware, the company behind the code, and their official partners.

Notable Features

  • Extensible Accounting: Allows integration of job-related measurements into the JSON-based accounting backend, visible in qacct -j.

  • Speed and Scalability: Features an improved multi-threaded architecture of the qmaster process to handle more jobs and qstat calls than ever.

  • Transition from C to C++: Simplifies the codebase, improves performance, and reduces complexities and errors.

  • Supplementary UNIX Group ID Support: Provides fine-grained access control.

  • New GPU Support: Introduces a new tool called qgpu, which simplifies the setup of GPU-based clusters and provides new features such as per-job GPU usage accounting (including power consumption, watts, errors, etc. in the qacct -j accounting records).

  • Support for NVIDIA's New ARM-Based Architectures: Includes support for architectures such as Grace Hopper and its successors.

  • Long Term Supported Releases: Compatible with a wide range of operating systems and hardware architectures (AMD64, ARM, RISCV, PPC, etc.). It also includes legacy OS support and more.