Three steps
From one phone to traceless compute.
A growing cloud bill
I was looking for a way to cut our Google Cloud bill.
A supercomputer in my pocket
My Pixel 8 Pro has a 9-core CPU and 12 GB of RAM. So I ran a cloud job on it.
It worked
A real service ran on a phone. Then three problems showed up.
Every job had to be installed
Each job put its code on the phone’s disk first.
Logs and files piled up
Data, logs and temp files filled the phone’s storage.
Privacy and security, both ways
Someone else’s code could read or write my phone’s storage, and anyone hosting my job could see what it runs.
No local storage, only compute
The service binary is fetched straight into RAM. 0 bytes written or read on the local disk.
Only the customer’s storage
The service reads and writes only through the customer’s API. Audit, trace and observability logs go back to customer storage.
Owners in full control
Device owners opt in, and decide when and how much their devices share.