Making Data Centers Greener by Fixing Wasteful Computer Systems
Massive data centers consume vast amounts of energy and strain electrical grids. An MIT researcher uses machine learning to make these computer networks more efficient and reduce waste.
In short: Massive data centers consume vast amounts of energy and strain electrical grids. An MIT researcher uses machine learning to make these computer networks more efficient and reduce waste.
Every time you stream music or join a video call, distant warehouses filled with humming computers spring into action, but these power-hungry giants often waste a staggering amount of energy.
What happened, in plain words
Christina Delimitrou, an associate professor at the Massachusetts Institute of Technology, is redesigning how massive data centers and cloud computing systems operate. By applying machine learning, her team helps computers run more efficiently, handles shared hardware better, and fixes application errors to reduce wasted energy and avoid building new power plants.
Key points
- Wasted capacity: Many large computing systems run at only about 15 percent capacity despite high user demand, leading to inefficient energy use.
- Machine learning solutions: Delimitrou uses machine learning to automate resource management in the cloud, helping systems squeeze more power out of existing hardware.
- Preventing application slowdowns: Her group developed a tool called Seer, which uses deep learning to anticipate and prevent problems in web applications before they cause widespread slowdowns.
- Mimicking closed systems: Because tech companies use private hardware and software that academic teams cannot access, her group builds clones like a tool called Ditto to study system performance.
Terms explained
- Machine learning — A type of computer technology that allows software to learn from data and make decisions without being explicitly programmed for every step. Example: A music app that learns your favorite songs and automatically suggests similar tracks you might enjoy.
- Cloud computing — Using a network of remote servers hosted on the internet to store, manage, and process data, rather than using a local computer. Example: Saving your photos to an online photo storage service so you can view them from any device.
- Deep learning — An advanced branch of machine learning that uses complex digital neural networks to solve difficult problems like recognizing speech or images. Example: A smartphone app that identifies different types of plants from a photograph you take.
Why it matters
Making cloud systems more efficient can help reduce the heavy electrical grid strain caused by growing data centers, while also giving everyday smartphone users more predictable app performance.
What we still don't know
Tech companies use private hardware and software that academic teams cannot access, meaning laboratory solutions might not always work in the real world.
Based on reporting from MIT News. This is an independent explainer, written in our own words with AI assistance; MIT News has not reviewed or endorsed it. Read the original for the full details.