Infrastructure for MLOps
Dayne Sorvisto · Apress eBooks · 2023
This chapter is about infrastructure. You might think of buildings and roads when you hear the word infrastructure, but in MLOps, infrastructure refers to the most fundamental services we need to build more complex systems like training, inference, and model deployment pipelines. For example, we need a way to create data stores that can store features for model training and servers with compute and memory resources for hosting training pipelines. In the next section, we will look at a way we can simplify the process of creating infrastructure by using containers to package up software that can easily be maintained, deployed, and reproduced.