rvirt-gpu

Virtualization for GPU that allows you to run local GPU apps and the code is actually run in the cloud, keeping your data local, without changing your app, at least not much.

Warning

This is still under development. Please do not use it with sensitive data for now, please wait for a stable release.
It's mostly ideal for experimental and learning projects.

Functionality

• vGPU is a virtualization layer for a GPU
• your local app "runs" on local vGPU
• local app decrypts the actual local data and sends the (CUDA) instructions to the remote GPU-Coordinator
• GPU-Coordinator distribute the instructions to multiple real GPUs
• then it sends the results back to vGPU which sends them to the local app

Three advantages emerge:

• You can execute AI models without having a local GPU
• Execute on multiple GPUs, in the end, for some use cases, it could be faster than running on a local GPU
• Execute ML models without storing your data on untrusted providers. It will keep some blocks in memory which are needed for executing current instructions. But at least they are obfuscated and protected by OS not to be easily copied by other processes from memory. Data in transit is secured by TLS

Cons

• I know it will be slow, but in cases where the data flow is small compared to processing time it could be a reasonable compromise for the security it gives you
• Also because the instructions are distributed to multiple GPUs, when possible, it could offer a better performance, in some cases, than locally

Schema

Wiki

Wiki

Credits

The main idea is of Andrei Mărcuţ.