
  <rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
    <channel>
      <title>Chaos and Order</title>
      <link>https://www.youngju.dev/blog</link>
      <description>천천히 올바르게. AI Researcher &amp; DevOps Engineer Youngju&#39;s blog. GPU/CUDA, LLM, MLOps, Kubernetes AI workloads, and data engineering — plus mindset essays on confidence, routines, health, and sport psychology.</description>
      <language>ko</language>
      <managingEditor>fjvbn2003@gmail.com (Youngju Kim)</managingEditor>
      <webMaster>fjvbn2003@gmail.com (Youngju Kim)</webMaster>
      <lastBuildDate>Wed, 12 Aug 2026 00:00:00 GMT</lastBuildDate>
      <atom:link href="https://www.youngju.dev/tags/time-slicing/feed.xml" rel="self" type="application/rss+xml"/>
      
  <item>
    <guid>https://www.youngju.dev/blog/gpu-cuda/2026-08-12-gpuops-mig-and-timeslicing.en</guid>
    <title>MIG and Time-Slicing — Two Ways to Share One GPU</title>
    <link>https://www.youngju.dev/blog/gpu-cuda/2026-08-12-gpuops-mig-and-timeslicing.en</link>
    <description>There are broadly two ways to put multiple workloads on one GPU: time-slicing, which divides time, and MIG, which divides hardware. Despite sounding similar, their isolation guarantees are nothing alike. This post works from the official NVIDIA GPU Operator documentation to lay out the configuration structure, node labels, advertised resource names, and hardware requirements for each, then explains what the absence of memory and fault isolation between time-sliced replicas actually means in production and how to decide which one a given workload needs. Third post in the Kubernetes GPU operations and observability series.</description>
    <pubDate>Wed, 12 Aug 2026 00:00:00 GMT</pubDate>
    <author>fjvbn2003@gmail.com (Youngju Kim)</author>
    <category>gpu</category><category>kubernetes</category><category>mig</category><category>time-slicing</category><category>nvidia</category><category>gpu-operator</category>
  </item>

  <item>
    <guid>https://www.youngju.dev/blog/gpu-cuda/2026-08-12-gpuops-mig-and-timeslicing.ja</guid>
    <title>MIGとtime-slicing — GPU一枚を複数で使う二つの方法</title>
    <link>https://www.youngju.dev/blog/gpu-cuda/2026-08-12-gpuops-mig-and-timeslicing.ja</link>
    <description>GPU一枚に複数のワークロードを載せる方法は大きく二つあります。時間を分けるtime-slicingとハードウェアを分けるMIGですが、名前が似て見えるのとは裏腹に隔離の水準がまったく異なります。本記事ではNVIDIA GPU Operatorの公式ドキュメントを基準に、両方式の設定ファイル構造とノードラベル、広告されるリソース名、対応ハードウェア条件を整理し、time-slicingのレプリカ間にメモリ隔離も障害隔離も無いという事実が実務で何を意味するのか、どのワークロードに何を選ぶべきかを判断基準としてまとめます。Kubernetes GPU運用・可観測性シリーズの第三回です。</description>
    <pubDate>Wed, 12 Aug 2026 00:00:00 GMT</pubDate>
    <author>fjvbn2003@gmail.com (Youngju Kim)</author>
    <category>gpu</category><category>kubernetes</category><category>mig</category><category>time-slicing</category><category>nvidia</category><category>gpu-operator</category>
  </item>

  <item>
    <guid>https://www.youngju.dev/blog/gpu-cuda/2026-08-12-gpuops-mig-and-timeslicing</guid>
    <title>MIG와 time-slicing — GPU 한 장을 여럿이 쓰는 두 가지 방법</title>
    <link>https://www.youngju.dev/blog/gpu-cuda/2026-08-12-gpuops-mig-and-timeslicing</link>
    <description>GPU 한 장에 여러 워크로드를 올리는 방법은 크게 두 가지입니다. 시간을 나누는 time-slicing과 하드웨어를 나누는 MIG인데, 이름이 비슷해 보이는 것과 달리 격리 수준이 전혀 다릅니다. 이 글은 NVIDIA GPU Operator 공식 문서를 기준으로 두 방식의 설정 파일 구조와 노드 라벨, 광고되는 리소스 이름, 지원 하드웨어 조건을 정리하고, time-slicing 복제본 사이에 메모리와 장애 격리가 없다는 사실이 실무에서 무엇을 뜻하는지, 어떤 워크로드에 무엇을 골라야 하는지를 판단 기준으로 정리합니다. 쿠버네티스 GPU 운영·관측 시리즈의 세 번째 글입니다.</description>
    <pubDate>Wed, 12 Aug 2026 00:00:00 GMT</pubDate>
    <author>fjvbn2003@gmail.com (Youngju Kim)</author>
    <category>gpu</category><category>kubernetes</category><category>mig</category><category>time-slicing</category><category>nvidia</category><category>gpu-operator</category>
  </item>

    </channel>
  </rss>
