vLLM ships an experimental AFD plugin to split MoE attention and FFN serving
AccBalanced · x · 2026-07-25
- A new experimental vLLM AFD Plugin is being shipped by the vllm-project with support from Ascend & vLLM, StepFunai, AntGroup, and FastAFD.
- It implements Attention-FFN Disaggregation for MoE serving: attention and expert/FFN workloads are separated into different services so they can be scaled independently.
- The plugin keeps the same vLLM serving surface and avoids forking the stack.
- The post notes support for both NVIDIA GPUs and Ascend NPUs.
Related event: vLLM Releases Experimental AFD Plugin to Boost MoE Inference(3 posts)→
More from Infra
- Engram's random reads don't suit SSDs; CPU-memory over NVLink could serve all 72 GPUs — bookwormengr · 2026-09-11
- 80% of the DIY LLM inference hype posters have already quit — it's brutally hard systems work — abhijithneil · 2026-09-11
- Hugging Face's Ultra Scale Playbook: a free book on training LLMs on GPU clusters — mdancho84 · 2026-09-11
- Is inference latency becoming the biggest bottleneck for production AI agents? — Euphoric_Sea632 · 2026-09-11
- LLM Serving Metrics Thread: Why TPOT and Uptime Make or Break User Experience — abhijithneil · 2026-09-11
- PlanetScale launches sharded Postgres: 768 servers acting as one, 1PB scale — dhruv2038 · 2026-09-11