Internship – End-of-Studies Research Project

Context & Motivation

Modern distributed storage infrastructures are made up of thousands of hard disk drives (HDDs) operating continuously. While compute and network energy costs are increasingly well understood, HDD/SSD power consumption under real-world workloads remains difficult to measure and predict accurately. Understanding and modeling this consumption is critical for:

• Reducing the environmental footprint of large-scale storage systems
• Improving capacity planning and thermal management
• Enabling proactive power optimization in production environments

To date, no robust, interpretable multi-parameter model exists that can accurately estimate disk energy consumption from observable parameters in a production environment. This internship aims to fill that gap.

Your Mission

As part of our R&D team, you will design, build, and validate a multi-parameter model for estimating HDD/SSD power consumption. Your work will include:

1. State of the Art

Review existing literature and open-source projects related to HDD/SSD power modeling, thermal behavior, and energy measurement in storage systems (including tools such as PowerAPI).

2. Physical Modeling

Develop a physics-based multi-parameter model of temperature and power draw, capturing relationships between:

• Drive temperature
• Fan speed and airflow
• Read/write throughput and IOPS
• Idle vs. active state transitions

3. Interpretable Machine Learning Model

Design a multi-parameter ML model (e.g., gradient boosting, linear regression with feature engineering) that is both accurate and interpretable — enabling engineers to understand which parameters drive consumption under different workload profiles.

4. Measurement & Validation
Instrument real drives in a controlled laboratory environment and under
production-representative workloads to collect ground-truth measurements. Use this data to:
• Calibrate and validate the model
• Quantify model accuracy across workload types
• Identify parameters with the greatest predictive value

Expected Outcomes & Valorization

Depending on results, the work may be valorized through:
• A scientific publication or technical white paper
• Integration into the open-source PowerAPI project
• Direct integration into Scality's internal monitoring and capacity planning tooling

Technical Stack

• Python (primary language for data collection, modeling, and analysis)
• scikit-learn (ML modeling and evaluation)
• PowerAPI ecosystem (https://powerapi.org/)
• Linux system tooling for hardware instrumentation (smartctl, lm-sensors, etc.)
• Jupyter Notebooks for exploratory analysis and result visualization

Candidate Profile

• Final-year student in a Master's program or Engineering school (Bac+5)
• Strong interest in physical modeling and/or applied machine learning
• Comfortable working with real hardware and experimental data
• Autonomous, curious, and rigorous in your approach to problem-solving
• Able to communicate results clearly in written and spoken English

Work Environment

You will join a multicultural R&D team with colleagues across France, the US, and Asia. English is our primary working language. You will benefit from: 
• Mentorship from senior R&D engineers with expertise in distributed systems and performance engineering
• Access to real production-grade storage hardware and laboratory infrastructure
• A high-trust environment where your findings will directly influence engineering decision

Why Join Scality?

• Work on a concrete research problem with real-world industrial impact
• Contribute to open-source energy-efficiency tooling used beyond Scality
• Be part of a team building infrastructure trusted by Fortune 500 companies
• Potential to publish research or continue as a full-time engineer after the internship