GPU Checkpointing - Revision history

Radostin at 22:13, 10 November 2025

2025-11-10T22:13:13Z

Radostin at 22:11, 10 November 2025

2025-11-10T22:11:45Z

Radostin at 18:24, 29 October 2025

2025-10-29T18:24:28Z

Radostin at 18:19, 29 October 2025

2025-10-29T18:19:07Z

Radostin at 18:17, 29 October 2025

2025-10-29T18:17:08Z

Radostin at 18:16, 29 October 2025

2025-10-29T18:16:35Z

Radostin at 18:13, 29 October 2025

2025-10-29T18:13:32Z

Radostin at 17:41, 29 October 2025

2025-10-29T17:41:21Z

Radostin at 17:40, 29 October 2025

2025-10-29T17:40:40Z

Radostin at 17:01, 29 October 2025

2025-10-29T17:01:51Z

← Older revision		Revision as of 22:13, 10 November 2025
Line 1:		Line 1:
−	This page explains how CRIU handles GPU-accelerated workloads, what vendor components are needed (NVIDIA & AMD), and how to use them for processes and containers. For more information, see the [https://arxiv.org/abs/2502.16631 CRIUgpu paper].	+	This page explains how CRIU handles GPU-accelerated workloads, what vendor components are needed (NVIDIA & AMD), and how to use them for processes and containers. For more information, see the
		+	NVIDIA's [https://developer.nvidia.com/blog/checkpointing-cuda-applications-with-criu/ Checkpointing CUDA Applications with CRIU] blog post and the [https://arxiv.org/abs/2502.16631 CRIUgpu paper].

	== How CRIU integrates with GPU checkpointing mechanisms? ==		== How CRIU integrates with GPU checkpointing mechanisms? ==

@@ Line 7: / Line 7: @@
 === CUDA Plugin ===
-The checkpointing functionality for CUDA applications is enabled through a command-line utility called [http://github.com/NVIDIA/cuda-checkpoint cuda-checkpoint]. This utility allows to transparently checkpoint and restore the CUDA state of a running Linux process. The CUDA plugin integrates this external utility with CRIU to safely pause, checkpoint, and restore processes and containers that use NVIDIA GPUs. It detects whether <code>cuda-checkpoint</code> is available and whether the system has an NVIDIA GPU; if not, the plugin disables itself.
+The checkpointing functionality for CUDA applications is enabled through the [http://github.com/NVIDIA/cuda-checkpoint cuda-checkpoint] utility. This utility allows to transparently checkpoint and restore the CUDA state of a running Linux process. The CUDA plugin integrates this external utility with CRIU to safely pause, checkpoint, and restore processes and containers that use NVIDIA GPUs. It detects whether <code>cuda-checkpoint</code> is available and whether the system has an NVIDIA GPU; if not, the plugin disables itself.
 === AMDGPU Plugin ===

@@ Line 7: / Line 7: @@
 === CUDA Plugin ===
-The checkpointing functionality for CUDA applications is enabled through a command-line utility called [http://github.com/NVIDIA/cuda-checkpoint cuda-checkpoint]. This utility allows to transparently checkpoint and restore the GPU state of a running Linux process.
+The checkpointing functionality for CUDA applications is enabled through a command-line utility called [http://github.com/NVIDIA/cuda-checkpoint cuda-checkpoint]. This utility allows to transparently checkpoint and restore the CUDA state of a running Linux process. The CUDA plugin integrates this external utility with CRIU to safely pause, checkpoint, and restore processes and containers that use NVIDIA GPUs. It detects whether <code>cuda-checkpoint</code> is available and whether the system has an NVIDIA GPU; if not, the plugin disables itself.
 === AMDGPU Plugin ===

← Older revision		Revision as of 18:19, 29 October 2025
Line 1:		Line 1:
−	This page explains how CRIU handles GPU-accelerated workloads, what vendor components are needed (NVIDIA & AMD), and how to use them for processes and containers. For more information, ~~check out~~ the [https://arxiv.org/abs/2502.16631 CRIUgpu paper].	+	This page explains how CRIU handles GPU-accelerated workloads, what vendor components are needed (NVIDIA & AMD), and how to use them for processes and containers. For more information, see the [https://arxiv.org/abs/2502.16631 CRIUgpu paper].

	== How CRIU integrates with GPU checkpointing mechanisms? ==		== How CRIU integrates with GPU checkpointing mechanisms? ==

← Older revision		Revision as of 18:17, 29 October 2025
Line 1:		Line 1:
−	This page explains how CRIU handles GPU-accelerated workloads, what vendor components are needed (NVIDIA & AMD), and how to use them for processes and containers. For more information, check out the [https://arxiv.org/abs/2502.16631 CRIUgpu paper], NVIDIA's [https://developer.nvidia.com/blog/checkpointing-cuda-applications-with-criu/ blog post], and AMD conference talks at [https://lpc.events/event/11/contributions/891/ LPC] and [https://indico.freedesktop.org/event/1/contributions/18/ X.Org].	+	This page explains how CRIU handles GPU-accelerated workloads, what vendor components are needed (NVIDIA & AMD), and how to use them for processes and containers. For more information, check out the [https://arxiv.org/abs/2502.16631 CRIUgpu paper].

	== How CRIU integrates with GPU checkpointing mechanisms? ==		== How CRIU integrates with GPU checkpointing mechanisms? ==

← Older revision		Revision as of 18:16, 29 October 2025
Line 10:		Line 10:

	=== AMDGPU Plugin ===		=== AMDGPU Plugin ===
		+
		+	== Usage Examples ==
		+
		+	== Limitations ==

@@ Line 1: / Line 1: @@
 This page explains how CRIU handles GPU-accelerated workloads, what vendor components are needed (NVIDIA & AMD), and how to use them for processes and containers. For more information, check out the [https://arxiv.org/abs/2502.16631 CRIUgpu paper], NVIDIA's [https://developer.nvidia.com/blog/checkpointing-cuda-applications-with-criu/ blog post], and AMD conference talks at  [https://lpc.events/event/11/contributions/891/ LPC] and [https://indico.freedesktop.org/event/1/contributions/18/ X.Org].
-CRIU checkpoints Linux-kernel resources (e.g., memory, threads, files, sockets). GPU state such as device memory, contexts, and queues that lives outside normal process address space needs special handling, so CRIU relies on vendor-specific [[plugins]].
+== How CRIU integrates with GPU checkpointing mechanisms? ==
-== NVIDIA CUDA ==
+CRIU checkpoints Linux-kernel resources (e.g., memory, threads, files, sockets). GPU state such as device memory, contexts, and queues that lives outside normal process address space needs special handling, so CRIU relies on vendor-specific [[plugins]].
-The checkpointing functionality for CUDA is enabled through a command-line utility called [http://github.com/NVIDIA/cuda-checkpoint cuda-checkpoint]. This utility allows to transparently checkpoint and restore the GPU state of a running Linux process.
+=== CUDA Plugin ===
-== AMD ROCm ==
+The checkpointing functionality for CUDA applications is enabled through a command-line utility called [http://github.com/NVIDIA/cuda-checkpoint cuda-checkpoint]. This utility allows to transparently checkpoint and restore the GPU state of a running Linux process.

← Older revision		Revision as of 17:41, 29 October 2025
Line 1:		Line 1:
−	This page explains how CRIU handles GPU-accelerated workloads, what vendor components are needed (NVIDIA & AMD), and how to use them for processes and containers. For more information, check out the [https://arxiv.org/abs/2502.16631 CRIUgpu paper], NVIDIA's [https://developer.nvidia.com/blog/checkpointing-cuda-applications-with-criu/ blog post], and AMD's [https://lpc.events/event/11/contributions/891/ LPC] and [https://indico.freedesktop.org/event/1/contributions/18/ X.Org] ~~conference talks~~.	+	This page explains how CRIU handles GPU-accelerated workloads, what vendor components are needed (NVIDIA & AMD), and how to use them for processes and containers. For more information, check out the [https://arxiv.org/abs/2502.16631 CRIUgpu paper], NVIDIA's [https://developer.nvidia.com/blog/checkpointing-cuda-applications-with-criu/ blog post], and AMD conference talks at [https://lpc.events/event/11/contributions/891/ LPC] and [https://indico.freedesktop.org/event/1/contributions/18/ X.Org].

	CRIU checkpoints Linux-kernel resources (e.g., memory, threads, files, sockets). GPU state such as device memory, contexts, and queues that lives outside normal process address space needs special handling, so CRIU relies on vendor-specific [[plugins]].		CRIU checkpoints Linux-kernel resources (e.g., memory, threads, files, sockets). GPU state such as device memory, contexts, and queues that lives outside normal process address space needs special handling, so CRIU relies on vendor-specific [[plugins]].

← Older revision		Revision as of 17:40, 29 October 2025
Line 1:		Line 1:
−	This page explains how CRIU handles GPU-accelerated workloads, what vendor components are needed (NVIDIA & AMD), and how to use them for processes and containers.	+	This page explains how CRIU handles GPU-accelerated workloads, what vendor components are needed (NVIDIA & AMD), and how to use them for processes and containers. For more information, check out the [https://arxiv.org/abs/2502.16631 CRIUgpu paper], NVIDIA's [https://developer.nvidia.com/blog/checkpointing-cuda-applications-with-criu/ blog post], and AMD's [https://lpc.events/event/11/contributions/891/ LPC] and [https://indico.freedesktop.org/event/1/contributions/18/ X.Org] conference talks.

	CRIU checkpoints Linux-kernel resources (e.g., memory, threads, files, sockets). GPU state such as device memory, contexts, and queues that lives outside normal process address space needs special handling, so CRIU relies on vendor-specific [[plugins]].		CRIU checkpoints Linux-kernel resources (e.g., memory, threads, files, sockets). GPU state such as device memory, contexts, and queues that lives outside normal process address space needs special handling, so CRIU relies on vendor-specific [[plugins]].