Memory and GPU Checkpointing (Beta)

Snapshot CPU and GPU memory to skip imports, model loading, and CUDA kernel compilation, cutting Cerebrium container cold start times.

Introduction

Memory checkpointing takes a snapshot of a container’s CPU memory and GPU memory, and uses it to speed up the startup of future containers. Applications that perform a large amount of work at container start time benefit the most from this process.

This is useful for both CPU-only and GPU workloads. For CPU applications, checkpointing can preserve expensive initialization work such as imports, dependency loading, configuration setup, and in-memory state. For GPU applications, it can also preserve model weights, CUDA state, and compiled kernels.

For example, ML and LLM frameworks often load large model weights and compile CUDA kernels at container start time, which can take many seconds or minutes. Loading from a checkpoint that already contains this initialized state can skip most of that delay.

Since this feature is still in beta, please report all issues to the team via our Discord Community or via Email.

How To Use

Checkpointing is available in early beta to our customer base. Add the following to your cerebrium.toml in order to use it.

[cerebrium.experimental]
checkpointing = true

To create a checkpoint your application has to send a trigger to our runtime after it has performed its initialization and is ready. When this trigger is received, the runtime verifies if a new checkpoint is required. To save resources, the system will not create a new checkpoint if:

1. A checkpoint already exists for the current build version.
2. Another container instance is already undergoing the checkpointing process.

If a checkpoint should occur your container will be frozen for the duration of the process. GPU memory will be copied to CPU memory and then all of the container memory will be written to storage. This saved checkpoint will then be distributed to be able to run throughout the region.

Send a POST request to http://169.254.169.253:8234/checkpoint from inside your container when the container is ready to checkpoint.

If successful subsequent containers will be restored from this created checkpoint. You can tell that a container was restored from a checkpoint if it has CEREBRIUM_RESTORED: container restored from checkpoint as the first log line in the container.

A checkpoint is tightly coupled to a single deployment. To disable restoring from checkpoints simply remove the POST request and redeploy your application.

You can find several implementations in our Examples repository on Github.

vLLM Example

from vllm import AsyncLLMEngine
from vllm.engine.arg_utils import AsyncEngineArgs
import http
import urllib

# Init vLLM engine
engine_args = AsyncEngineArgs(
    model="Qwen/Qwen2.5-0.5B-Instruct",
    async_scheduling=False,
    sleep_mode=True
)
engine = AsyncLLMEngine.from_engine_args(engine_args)

# Drop KV cache for reduced GPU memory footprint.
engine.sleep(level=1)
# Trigger checkpoint
try:
    import json
    req = urllib.request.Request("http://169.254.169.253:8234/checkpoint", method="POST")
    with urllib.request.urlopen(req, timeout=300) as response:
        result = response.read()
        print(json.loads(result))
except http.client.RemoteDisconnected:
    # TCP connections disconnect on restore and throw remote
    pass

# Restore KV cache
engine.wake_up()

Limitations

Memory Overhead: The container memory allocation needs to be large enough to contain the GPU memory dump in addition to your regular memory use.

Execution Lifecycle: When a container is restored from a checkpoint execution continues from the point where the http request is sent. If environment variables were read before this point they will remain the same as they were from the time of the checkpoint.

Network Connections: Any TCP connections that were made before the checkpoint will have disconnected. For example if you connected to a database before the checkpoint you will have to reestablish that connection after restore.

Ephemeral Filesystem: Any files written to disk before the checkpoint will not be copied to the restored container. Only memory is checkpointed.

Provider Availability: Checkpointing is only available on the AWS provider. More coming soon.

Platform specific recommendations

vLLM

vLLM checkpointing support is not complete but still possible. See https://github.com/vllm-project/vllm/issues/34303 and other issues.

If you are getting an EngineCoreDead exception add async_scheduling=False to your AsyncEngineArgs and it should succeed.

The larger the size of the memory checkpoint the slower the restore is. We can reduce the size of the snapshot substantially and improve startup times by dropping the KV Cache before checkpoint and recreating it after restore. vLLM has functionality that does this built in as part of vLLM Sleep Mode.