Everpure Partners with University of Washington to Advance Cloud Efficiency Research
Everpure is partnering with the University of Washington’s Center for the Future of Cloud Infrastructure (FOCI), working with faculty and students to explore techniques for building more scalable, more efficient large-scale cloud-based systems. While FOCI is investigating many areas in cloud systems, Everpure has a particular interest in increasing the efficiency of storage systems for large-scale cloud and AI/HPC infrastructure.
Why Storage Efficiency Matters
While storage systems traditionally consume a relatively small fraction of the power used by data centers, they still represent a large amount of power usage, and that amount is
increasing as users demand ever-larger and ever-faster storage systems. Everpure DirectFlash® Modules (DFMs) currently consume up to 30 watts in active use, meaning an exabyte-scale system with 4,000 DFMs would consume 120 kilowatts. With servers and networking, a complete system may consume upwards of 200 kilowatts.
Tackling Peak and Total Power Consumption
A major goal of the research with the University of Washington is to develop approaches that reduce both overall power consumption and peak power consumption. Reducing peak power consumption, even without reducing total energy usage (energy is power over time), is particularly important, since it allows storage systems to use smaller power supplies and run in smaller power envelopes, reducing the need to run higher-capacity electric feeder lines to data centers.
Reducing peak power consumption also results in a more efficient data center by reducing peak cooling requirements and allowing systems to reduce power lost to cooling and power conversion. Reducing overall power consumption is also a goal, since it results in more energy-efficient storage systems and reduces the costs of running a data center. Lower overall power consumption also further increases the energy advantage of flash-based storage systems over hard drive-based storage systems.
Our support builds on FOCI’s broader research agenda on energy-aware cloud infrastructure. FOCI researchers are exploring how to operate large-scale systems within tight and dynamic power budgets, designing mechanisms that make compute, memory, and storage more power-adaptive while preserving performance and reliability. We will bring this expertise to bear on flash-based storage, co-designing hardware and software that explicitly optimizes for performance per watt and efficiency under real-world power constraints.
Designing for Energy-aware Performance
In addition to better managing power consumption, we hope to find new ways to leverage DirectFlash technology to balance performance, density, and energy consumption for demanding workloads. This endeavor directly addresses the interdependent challenges of power efficiency and physical space utilization. As AI and HPC infrastructures scale, the associated operational expenditures (OPEX) for power, cooling, and data center rack space become critical constraints.
Our hypothesis is that the fine-grained control afforded by DirectFlash can enable novel system-level co-design strategies. This FOCI collaboration will allow us to investigate techniques that optimize not just for raw IOPS or capacity, but for performance per watt and density per rack unit, ultimately maximizing overall infrastructure efficiency.
Everpure + UW FOCI: A Shared Vision for Sustainable Infrastructure
We are excited to embark on this research journey with the University of Washington. Collaboration with FOCI provides an opportunity to rethink storage system design for power-constrained, high-density environments. We believe this work will contribute valuable, generalizable principles to the research community, helping to define the future of efficient cloud and AI infrastructure. We look forward to sharing our findings as the research progresses and to integrating the findings into Everpure products to improve their capabilities.