# UCSD CSE 291A lecture 13 covers prefix caching and P–D disaggregation

> Original title: UCSD CSE 291A 第 13 讲：Prefix Caching 与 Prefill–Decode 解耦，分布式 LLM 推理怎样少做重复计算

- Source: meng shao
- Published: 2026-11-10T00:40:40.000Z
- Bad Influence: https://badinfluence.news/items/uk905tbhsul8hi7izpneit4et
- Original: https://x.com/shao__meng/status/2119953352901755240

## Summary

UCSD CSE 291A Lecture 13, taught by Yiying Zhang, covers prefix caching and Prefill–Decode disaggregation for distributed LLM serving. The main reading is Preble (ICLR 2025), whose global E2 scheduler and local fair scheduler report 1.5–14.5x lower average latency than SGLang in the lecture's own results. The lecture also covers DistServe, which splits prefill and decode workers and reports about 3.3 requests/s per GPU in an illustrative example.

## Post

https://x.com/i/article/2119948219887341568