create a research document on the current theory, tools and providers around distributed GPU compute

[NiklasTR]

Problem

We are not the only ones thinking about ways to use decentralized hardware to perform usefule computational tasks. The simplest version of a $LAB miner is a repurposed ETH-miner turned into a GPU workstation. The workstation functions as a small server that is running a LabDAO approved container and exposing an API gateway endpoint within the openlab network. Users submit requests to the API and jobs are processed once payments are submitted to an escrow contract by the user. In the most complex stage, a group of ETH-miners / a mining-pool come together and develop networking tools to enable high-throughput dynamic GPU-based inference services. Just like in the previous example an API gateway is receiving requests, but now relays requests to a cluster of GPU-powered machines.

Solution

Let's put together a research document covering these tasks:

• collect theory, patterns, designs, and tools/services that already solve parts of the "massively distributed GPU compute" problem
• collect data and case studies for understanding the market, including:
  • different potential untapped provider groups (miners, gamers, labs, free cloud credits)
  • companies operating in the space, from infrastructure through to saas

The document is a markdown file that is added to this repository.