DDN Becomes the First Storage Vendor Natively Integrated Into NVIDIA KV Cache Management
DDN Blog, Thursday, June 25th, 2026
DDN's Infinia plugin ships natively in NVIDIA's NIXL package, simplifying KV cache acceleration for inference.
DDN became the first storage vendor whose plugin is included in NVIDIA's official NIXL inference transfer library, starting with version 1.3 in July 2026.
The DDN Infinia NIXL plugin ships directly in NVIDIA's standard package and prebuilt Dynamo inferencing container, letting customers activate KV cache acceleration via configuration without code changes.
It provides zero-copy transfers, parallel batch execution, asynchronous coroutine APIs, and multi-tenant isolation.
It works with vLLM, SGLang, LMCache, and TensorRT-LLM.
The integration reduces engineering overhead and prompted NVIDIA to upgrade NIXL to C++20.