Logo
Sign in
Product Logo
Power ManagementAdaptive Computing

Workload-aware power management for HPC systems that reduces energy costs, carbon footprint, and unnecessary power consumption.

Vendor

Vendor

Adaptive Computing

Company Website

Company Website

Intelligent-Pow…-Paper_20191001.pdf
Power-Mana…t_20190930.pdf
Product details

Power Management is a workload-aware solution integrated into the Moab® HPC Suite, designed to optimize energy usage across high-performance computing (HPC) environments. With energy costs often equaling hardware costs over a system’s lifetime, this solution helps organizations reduce expenses, meet power targets, and minimize environmental impact. It achieves this through intelligent idle system power reclamation and per-application CPU clock frequency optimization.

Features

Idle System Power Reclamation

  • Automatically lowers power state of idle nodes (standby, suspend, hibernate, or off).
  • Identifies idle periods such as evenings, weekends, holidays, or low job submission times.
  • Utilizes a Green Pool Buffer Policy to maintain responsiveness by keeping a small pool of idle nodes online.
  • Predictive power management ensures readiness for new jobs while saving energy. Per-Application Power Optimization
  • Automates CPU clock frequency adjustments based on job type.
  • Allows users to override defaults at job submission to set specific clock speeds.
  • Reduces clock speed for non-CPU intensive or low-priority workloads.
  • Admins can assign clock settings to Moab® job templates for consistent application.

Benefits

  • Eliminate Excess Power Consumption: Automatically powers down unused nodes to save energy.
  • Decrease Carbon Footprint: Reduces environmental impact through smarter energy use.
  • Maintain System Responsiveness: Green Pool ensures quick access to powered nodes.
  • Reduce Unnecessary Power Usage: Tailors CPU performance to workload needs.
  • Minimize Energy Costs: Cuts operational expenses by optimizing power consumption.
  • Stay Within Power Caps: Ensures workloads operate within application-optimized energy limits.