We Need Better Interfaces for Programming the Internet of Things

TL;DR

Hardware specialization is disrupting the software industry, leading to the need for new programming abstractions and techniques for optimizing performance and energy efficiency.

Transcript

I'm Katherine McKinley I'm a principal researcher at Microsoft in Redmond my research area is programming language implementation I tried to say make software better faster more fish energy-efficient higher performance and correct hardware is in a very disruptive period the end of Dennard scaling which means that hardware is you know here hardware ... Read More

Key Insights

• 👶 Hardware specialization, such as FPGAs and combined processors, is disrupting the software industry and requires new programming techniques.
• 🥺 Optimizing for both average and tail latency leads to a better user experience, requiring heterogeneous hardware.
• ❓ Programmability is crucial for innovation in programming language and hardware.
• 🪛 Applications drive innovation in programming language and hardware.

Questions & Answers

Q: How is hardware disruption affecting the software industry?

Hardware disruption, characterized by specialization and a move away from Dennard scaling, is forcing software systems to adapt to different types of hardware. This requires new programming techniques and abstractions to ensure software can run efficiently and effectively on various generations of hardware.

Q: What is Katherine McKinley's research area at Microsoft?

Katherine McKinley is a principal researcher at Microsoft in the field of programming language implementation. Her research focuses on making software better, faster, more energy-efficient, and high-performance, especially during this disruptive period in hardware.

Q: What are some examples of hardware specialization mentioned by McKinley?

McKinley mentions FPGAs (Field-Programmable Gate Arrays) and the combination of big and little processors as examples of hardware specialization. These specialized processors offer different levels of power and energy efficiency, allowing for better optimization in software.

Q: How does McKinley propose handling inaccuracies in sensor data in programming models?

McKinley aims to develop programming abstractions that can handle inaccuracies in sensor data without requiring complex statistical or computer science knowledge. These abstractions will allow developers to reason about the fact that sensor data is not always correct, making it easier to program applications that rely on sensor inputs.

Summary & Key Takeaways

• Katherine McKinley, a principal researcher at Microsoft, discusses the disruptive period in hardware and the challenges it poses for software systems.

• Hardware specialization, such as FPGAs and combined big and little processors, requires software to adapt and be able to run on different generations of hardware.

• McKinley's research focuses on developing programming abstractions that balance quality and energy efficiency and handle inaccuracies in sensor data.