#113. 缓存——“计算机科学的两大难题之一“

16 Mar 2023 · 85 min
牛油果烤面包

16 Mar 2023 · 85 min

0:00

In this episode, hosts 斯科亚特 and Wendy interview cache expert 岳瑶, co‑founder of IOPSystems and creator of the open‑source PelicanCache framework. The conversation explores the fundamental challenges of caching—balancing freshness, cost, and latency—through concrete mechanisms such as dual TTL (an insurance period plus a short validation window) and proactive push‑based invalidation. They discuss the architecture of distributed cache services, emphasizing the need for high‑performance networking and the impact of RPC latency when thousands of cache calls are made per user request. Language choices are examined, with a preference for low‑level compiled languages (C, Go) over higher‑level runtimes to meet strict latency budgets. Looking forward, they identify two major research directions: richer data‑type support and API‑driven cache capabilities, and the integration of memory and flash storage (e.g., CXL‑based pools) to create hybrid caching layers. The episode concludes with practical advice for engineers entering the field, recommending study of classic C‑based cache codebases (Memcached, Redis) and participation in the PelicanCache community on Discord.

Chapters

More from Cat

43:35
Podcast

#152. 不做 AI 硬件,做 AGI 配件:一枚 AI 戒指背后的工程卓越

3mo ago
1:12:59
Podcast

#151. Hello again, Apple Intelligence? 唠唠 WWDC26

4mo ago
55:00
Podcast

#150. 如何从零到一做风投

4mo ago
1:21:06
Podcast

#149. 量子计算和量子计算机

5mo ago
1:37:54
Podcast

#148. 我们为什么AI焦虑,我们又该怎么脚踏实地

7mo ago
1:22:19
Podcast

#147. 圆桌:AI如何改变了IT公司的工作方式

8mo ago
1:30:56
Podcast

#146. 2025年终感悟与推荐

9mo ago
58:15
Podcast

#145 AI Agent - 智能体

10mo ago
1:12:17
Podcast

#144. 大模型时代的AI该如何以人为本

11mo ago
1:01:33
Podcast

#143. 规模、质量、定制化——GenAI时代如何做产品?

11mo ago
1:24:04
Podcast

#142. 基因编辑——从细菌的防御武器到基因手术刀

1y ago
See all content