– Repurposing decommissioned smartphones reduces embodied carbon from manufacturing.
– Clusters of 25-50 smartphone motherboards match traditional server performance with lower energy.
– Kubernetes and a specialized Linux distribution manage the smartphone-based cloud infrastructure.
– A 20-board cluster can handle a lecture of 75+ students with low latency.
– The final 2,000-device system will support ~100 courses, equaling about 50 conventional servers.
A current collaborative project between the University of California San Diego (UCSD) and Google aims to address the topic of sustainability in information technology. The researchers hope to develop a method that repurposes discarded smartphones as resource-efficient cloud computing platforms. This approach is designed to reduce “embodied carbon,” namely the greenhouse gas emissions generated during the production and transportation of electronic components. To build the server cluster, the scientists are removing the motherboards from approximately 2,000 decommissioned Google Pixel smartphones.
How the server clusters actually work now
The motherboards will be combined into clusters and managed using container tools such as Kubernetes. Instead of the Android mobile operating system, a specially adapted Linux distribution will be used. A physical cluster will consist of 25 to 50 smartphone motherboards, which is expected to achieve the computing power of a conventional server mainframe, but with lower energy consumption during operation. This is a pretty smart way to reuse old electronics that would otherwise just end up in a landfill somewhere.
Testing and real world performence seen so far
The planned data center at UCSD is intended to provide the university with a cost-effective and low-emission cloud computing infrastructure. Tests have already shown that a network of just 20 smartphone boards is sufficient to handle the requests of a lecture with over 75 students stably and with low latency. The final system, with 2,000 integrated devices, is expected to be capable of supporting around 100 such courses simultaneously and offering computing capacity equivalent to approximately 50 traditional servers. That’s alot of processing power for basically free hardware.
- Pricing and specification details: Not specifically given, but using 2,000 discarded Google Pixel motherboards implies near-zero hardware cost for the boards themselves.
- Energy consumption: Lower than conventional server mainframes during operation, making it a greener option for cloud computing.
- Performance: 20 boards handled 75+ students with low latency; 2,000 boards equals about 50 traditional servers worth of capacity.
