Energy costs for datacenters are increasing and become a substantial portion of data centers' total cost of ownership (TCO). Specifically, the datacenter storage system is quickly expanding to accommodate newly generated digital data. Hence, reducing the energy consumption of IT systems is an important challenge. To date, energy aware management for storage systems has been limited. Our goal is to take significant steps towards energy efficient storage management.
Our activities include:
- Study the factors that affect power usage in storage systems
- Develop models, metrics, and benchmarks that enable the calculation and management of energy resources and energy efficiency
- Develop means to control the energy aspect of storage
- Build prototype systems that implement the models and means in a real world setting.
In our work, we are targeting all relevant scales of storage, from the low end, through SMBs, to large archives, and enterprise datacenters. We are also investigating ways to expose power consumption details to the application layer to enable application level energy aware management, and to other analysis tools like data mining and software optimizers. We attempt to identify the power consumption components for storage controllers and provide trade-offs between performance and power.
- SAMCEE (Storage Advanced Manager for Energy Efficiency) and GAMES (Green Active Management of Energy in IT Service Centres) – the most recent activity for building energy aware storage management as part of an EU datacenter management project
- Usage Centric Power Metrics – novel usage centric energy efficiency metrics for storage
- Storage Power Modeling – work that formed the foundation for power consumption estimation for storage disks, volumes and applications
- Disk Acoustic Modes – earlier work that characterized the workloads for which HDD acoustic quiet mode is energetically beneficial. The insights gained were leveraged in the GAMES/SAMCEE work
- Temperature aware data placement – work that attempted to save storage power and energy by considering temperatures, which might trigger energy costly cooling actions.