<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>resource defragmentation on Edwardesire</title><link>/tags/resource-defragmentation/</link><description>Recent content in resource defragmentation on Edwardesire</description><generator>Hugo -- gohugo.io</generator><language>en</language><copyright>&amp;copy; 2021 &lt;a href="https://www.edwardesire.com/">Edward Desire&lt;/a></copyright><lastBuildDate>Fri, 31 Jul 2026 11:20:00 +0800</lastBuildDate><atom:link href="/tags/resource-defragmentation/index.xml" rel="self" type="application/rss+xml"/><item><title>从空闲GPU到有效算力：大规模AI集群调度的演进</title><link>/posts/from-idle-gpu-to-effective-compute/</link><pubDate>Fri, 31 Jul 2026 11:20:00 +0800</pubDate><guid>/posts/from-idle-gpu-to-effective-compute/</guid><description>随着大模型训练和推理业务快速增长，GPU集群规模从千卡、万卡逐渐走向十万卡规模。表面上看，GPU资源越来越丰富，但生产环境中一个新的矛盾逐渐显现： 物理</description></item></channel></rss>