kvcached logo

Python Engines Blog arXiv: GPU OS vision
arXiv: Multi LLM Serving Slack Join DeepWiki Homepage License

Make GPU Sharing Flexible and Easy

Make GPU Sharing Flexible and Easy

kvcached (KV cache daemon) is a KV cache library for LLM serving/training on **shared GPUs**. By bringing OS-style **virtual memory** abstraction to LLM systems, it enables **elastic and demand-driven** KV cache allocation, improving GPU utilization under dynamic workloads. kvcached achieves this by decoupling GPU virtual addressing from physical memory allocation for KV caches. It allows serving engines to initially reserve virtual memory only and later back it with physical GPU memory when the cache is actively used. This decoupling enables on-demand allocation and flexible sharing, bringing better GPU memory utilization under dynamic and mixed workloads. Check out more details in the [blog](https://yifanqiao.notion.site/Solve-the-GPU-Cost-Crisis-with-kvcached-289da9d1f4d68034b17bf2774201b141).

Key Features

- **Elastic KV cache**: allocate and reclaim KV memory dynamically to match live load. - **GPU virtual memory**: decouple logical KV from physical GPU memory via runtime mapping. - **Memory control**: enforce memory limits with the kvcached CLI or the [revisioned production-control API](examples/02_memory_control#embedding-revisioned-limits-in-a-production-controller). - **Frontend router and sleep mode**: route requests to the target models and put models to sleep when idle. - **Support mainstream serving engines**: integrate with SGLang and vLLM. - **Prefix caching**: support automatic prefix caching (APC) with a configurable memory bound. See [the example doc](examples/09_prefix_caching) for details. ## 📢 Updates - **[2026-10]** kvcached **v0.1.6** supports **vLLM v0.30** (including Model Runner V2) and **SGLang v0.5.20**, hybrid linear-attention models (Qwen3.5, Qwen3.8), Gemma 4, and AMD GPUs (ROCm). - **[2026-04]** kvcached is **featured by Red Hat** for running LLMs dynamically in production under limited resources! Red Hat's [Sardeenz](https://github.com/rh-aiservices-bu/sardeenz) builds on kvcached to provide dynamic multi-model serving with Kubernetes and OpenShift support. See the [blog post](https://www.redhat.com/en/blog/running-llms-dynamically-production-limited-resources-hard-we-think-theres-room-another-approach) for more details. [[▶ View Demo]](https://app.arcade.software/share/xZoDfo1vyDbZrbZTK2gv?ref=share-link) - **[2026-04]** Added **prefix caching** support. kvcached now supports **automatic prefix caching (APC)** for vLLM and **RadixCache** for SGLang, enabling cross-request prefix reuse while maintaining elastic memory management. - **[2026-03]** Added **pipeline parallelism** support. MLA models (DeepSeek-V3, DeepSeek-V2 etc.) and GPT-OSS hybrid attention models (`openai/gpt-oss-20b`) are now also supported in **vLLM**. GPT-OSS support in SGLang updated to **v0.5.9**. ### Supported engines and models | Engine | Versions | Attention types | Example models | |--------|----------|-----------------|----------------| | SGLang | ≥ v0.5.11 (tested up to v0.5.20) | MHA / GQA / MLA / sliding window / hybrid | DeepSeek-V3, Qwen3-8B, GPT-OSS-20B, Qwen3.5-9B, Qwen3.8-27B, Gemma-4-E2B-it | | vLLM | ≥ v0.17.0 (tested up to v0.30.0) | MHA / GQA / MLA / sliding window / hybrid | DeepSeek-V3, Qwen3-8B, GPT-OSS-20B, Qwen3.5-9B, Qwen3.8-27B, Gemma-4-E2B-it, Gemma-4-12B-it | The minimum versions follow from the PyTorch requirement below. See [#425](https://github.com/ovg-project/kvcached/issues/425) for per-model results on each engine and KV layout. ## Example use cases
Multi‑LLM serving Multi‑LLM serving
kvcached allows multiple LLMs to share a GPU's memory elastically, enabling concurrent deployment without the rigid memory partitioning used today. This improves GPU utilization and saves serving costs.
Serverless LLM Serverless LLM
By allocating KV cache only when needed, kvcached supports serverless deployments where models can spin up and down on demand.
Compound AI systems Compound AI systems
kvcached makes compound AI systems practical on limited hardware by elastically allocating memory across specialized models in a pipeline (e.g., retrieval, reasoning, and summarization).
GPU workload colocation GPU workload colocation
kvcached allows LLM inference to coexist with other GPU workloads such as training jobs, fine-tuning, or vision models.
See concrete examples here: [kvcached/examples](https://github.com/ovg-project/kvcached/tree/main/examples). ## kvcached in action The following simple example shows how kvcached enables an unmodified vLLM engine run with dynamically allocated memory.

kvcached in action

## Performance: Multi-LLM serving kvcached enables dynamic memory sharing between LLMs, allowing them to share the same GPU memory elastically. As a comparison, the current serving engines need to statically reserve GPU memory at startup. This benchmark shows the performance benefits of kvcached when serving three `Llama-3.1-8B` models on an A100-80G GPU under workloads with intermittent peaks. kvcached can achieve **2-28x TTFT reduction** compared to the current serving engines. This performance gain can be converted to **significant cost savings** for LLM serving. Without kvcached, the systems have to provision more GPUs to achieve the same performance. Details can be found in [benchmarks/bench_latency_benefit](https://github.com/ovg-project/kvcached/tree/main/benchmarks/bench_latency_benefit).

TTFT mean TTFT p99

## Installation ### Prerequisites - Python (tested with 3.10 - 3.13) - PyTorch >= 2.10 (kvcached builds against the PyTorch stable ABI) - SGLang (tested with v0.5.20) or vLLM (tested with v0.30.0) kvcached can be installed as a plugin with existing SGLang or vLLM environment. ### Install from PyPI ```bash pip install kvcached --no-build-isolation ``` ### Install from source ```bash # under the project root folder pip install -e . --no-build-isolation --no-cache-dir ``` ### Using Docker kvcached installed with original engine dockers. ```bash docker pull ghcr.io/ovg-project/kvcached-sglang:latest # kvcached-v0.1.6-sglang-v0.5.20 docker pull ghcr.io/ovg-project/kvcached-vllm:latest # kvcached-v0.1.6-vllm-v0.30.0 ``` We prepare an all-in-one docker for developers: ```bash docker pull ghcr.io/ovg-project/kvcached-dev:latest ``` More instructions can be found [here](https://github.com/ovg-project/kvcached/blob/main/docker/README.md). ## Documentation kvcached is indexed on [DeepWiki](https://deepwiki.com/ovg-project/kvcached) for LLM-powered documentation. The documentation covers: - Core architecture and memory management system - Integration with vLLM and SGLang - Multi-model serving and controller system - Deployment guides and configuration reference - Performance benchmarking and analysis - Development tools and testing ## Testing kvcached can be enabled by setting the following environmental variables: ```bash export ENABLE_KVCACHED=true export KVCACHED_AUTOPATCH=1 ``` If you are using the engine-specific dockers, you can test kvcached by running the original engines' benchmark scripts. For example: ```bash # for sglang python -m sglang.launch_server --model meta-llama/Llama-3.2-1B-Instruct --port 30000 python -m sglang.bench_serving --backend sglang-oai --model meta-llama/Llama-3.2-1B-Instruct --dataset-name sharegpt --request-rate 10 --num-prompts 1000 --port 30000 # for vllm vllm serve meta-llama/Llama-3.2-1B-Instruct --port=12346 vllm bench serve --model meta-llama/Llama-3.2-1B-Instruct --request-rate 10 --num-prompts 1000 --port 12346 ``` > [!NOTE] > If you prefer to disable prefix caching, use `--no-enable-prefix-caching` for vLLM and `--disable-radix-cache` for SGLang. > > When kvcached is enabled, there is NO need to set memory utilization limit (e.g., using `--gpu-memory-utilization`) as kvcached will automatically manage the memory. > [!NOTE] > **AMD / ROCm:** on ROCm (HIP) builds, kvcached automatically defaults to the **non-contiguous** KV-cache layout. The contiguous layout (the default on NVIDIA) hands vLLM's ROCm attention backend strided per-layer KV tensors it cannot read correctly, whereas non-contiguous matches the layout the backend expects. You can override with `KVCACHED_CONTIGUOUS_LAYOUT=true|false`, but contiguous is not recommended on ROCm. If you installed kvcached using its source code, you can also do the following: ```bash cd benchmarks/simple_bench ./start_server.sh [sglang|vllm] --venv-path $VENV_PATH --model meta-llama/Llama-3.2-1B-Instruct # Wait until LLM server is ready ./start_client.sh [sglang|vllm] --venv-path $VENV_PATH --model meta-llama/Llama-3.2-1B-Instruct ``` The benchmark scripts automatically set `ENABLE_KVCACHED=true`. Please refer to each script for instructions on how to run inference with kvcached. > [!NOTE] > We haven’t fully tested kvcached with every version of SGLang and vLLM (there are too many!). If you run into issues with a specific version, please open an issue---we'll look into it and fix it within a few hours. ## Contributing We are grateful for and open to contributions and collaborations of any kind. We use pre-commit to ensure a consistent coding style. You can set it up by ``` pip install pre-commit pre-commit install ``` Before pushing your code, please run the following check and make sure your code passes all checks. ``` pre-commit run --all-files ``` ## Contacts kvcached is developed by many contributors from the community. The best way to contact us for questions, issues, and contributions, is through our [Slack channel](https://join.slack.com/t/ovg-project/shared_invite/zt-3fr01t8s7-ZtDhHSJQ00hcLHgwKx3Dmw) or [GitHub Issues](https://github.com/ovg-project/kvcached/issues). ## Trusted and deployed by

Google    LinkedIn    Intel    AMD    Red Hat    Adobe    Sony    ByteDance    Alibaba Cloud    Tencent    Moreh    Parasail

## Citation If you find kvcached useful, please cite our paper: ```bibtex @article{xing2025towards, title={Towards Efficient and Practical GPU Multitasking in the Era of LLM}, author={Xing, Jiarong and Qiao, Yifan and Mo, Simon and Cui, Xingqi and Sela, Gur-Eyal and Zhou, Yang and Gonzalez, Joseph and Stoica, Ion}, journal={arXiv preprint arXiv:2508.08448}, year={2025} } @article{yu2026prism, title={Prism: Cost-Efficient Multi-LLM Serving via GPU Memory Ballooning}, author={Yu, Shan and Qiao, Yifan and Ma, Mingyuan and Li, Yangmin and Yang, Shuo and Tong, Xinyuan and Wang, Yang and Xie, Zhiqiang and An, Yuwei and Cao, Shiyi and Bao, Ke and Vij, Deepak and Ding, Xiaoning and Wang, Yichen and Lu, Qingda and Wang, Zhong and Gao, Gao and Xu, Harry and Shu, Junyi and Xing, Jiarong and Sheng, Ying}, journal={OSDI}, year={2026} } ```