The more I look at LLM serving, the more it feels like the main object is not the request, the model, or even the GPU.

I became interested in LMCache because it sits in the part of LLM serving that feels both very practical and very under-discussed: KV cache movement.