PBS Professional's Green Provisioning: Power efficiency
PBS Professional’s Green Provisioning™ can monitor, shutdown and restart your HPC systems based on your requirements, ultimately managing power consumption to reduce energy usage. Validated by several large scale customers, Green Provisioning has lowered their energy use by up to 20 percent
Keeping high-performance computing (HPC) affordable and cost effective has always been key requirements. Recently, energy conservation and minimizing operating costs have become increasingly important factors shaping how HPC resources are used today.
Data centers are growing at a tremendous rate, but energy costs are quickly becoming a proportionally higher part of most IT operations budgets. Today, energy costs – to power and cool systems and data centers – consume the largest percentage of operational budgets.
Scalable and resilient to support thousands of compute nodes and limitless jobs, our Green Provisioning offering saves a significant percentage of power and cooling costs for your computer systems.
Main features and benefits of our Green Provisioning offering include:
- Dynamic system monitoring: Monitors the state of the queues and the level of activity on each node to decide if nodes running no jobs can be powered down or if queuing jobs require powered down nodes to be booted up.
- Customizable power reduction methods: The mechanism to adjust power use can be customized to suit the computing environment. Throttle CPU frequency to minimize idle power consumption, or fully shutdown nodes to eliminate consumption entirely.
- Node power down priorities: To ensure the highest performance nodes are available as much as possible, System Administrators have the ability to prioritize which nodes should be powered down first by assigning a “power down priority” to each node.
- Node power cycle time delays: Each node in the system has its own parameters that control how long the node is idle before it is powered down and how long to wait for the node to boot up.
- Temperature-based placement: Schedule jobs to nodes based on temperature to reduce cooling costs by evenly spreading out the A/C loads.
- Custom resources are supported: Fully supports node-level custom resources. When working out which powered down nodes should be booted up, the custom resources requested by each queued job are matched against the resources available on each node.
- Multi-node jobs are supported: Fully supports jobs that request multiple nodes or multiple chunks.
Please contact us
to learn more of our Green Provisioning’s offerings and energy benefits.