# Prefill-as-a-Service: KVCache of Next-Generation Models Could Go Cross-Datacenter

arXiv: https://arxiv.org/abs/2604.15039 · HTML: https://mlsanity.com/paper/2604.15039/

## Key facts

- Cited by 11 papers in the map (data through 14 Oct 2026).
- Placed in KV cache optimization, one of 12 directions in Efficient LLM inference and serving.
- Builds on 26 of its 34 references that are in the map.
- First posted on arXiv 16 Apr 2026.

## Placed in

- Efficient LLM inference and serving / [KV cache optimization](https://mlsanity.com/inference/kv-cache-optimization/)

## Builds on (most cited first, 26 of 26)

- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Longformer: The Long-Document Transformer](https://mlsanity.com/paper/2004.05150/) (2020)
- [Generating Long Sequences with Sparse Transformers](https://mlsanity.com/paper/1904.10509/) (2019)
- [gpt-oss-120b & gpt-oss-20b Model Card](https://mlsanity.com/paper/2508.10925/) (2025)
- [GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints](https://mlsanity.com/paper/2305.13245/) (2023)
- [DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model](https://mlsanity.com/paper/2405.04434/) (2024)
- [H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models](https://mlsanity.com/paper/2306.14048/) (2023)
- [KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cache](https://mlsanity.com/paper/2402.02750/) (2024)
- [DistServe: Disaggregating Prefill and Decoding for Goodput-optimized Large Language Model Serving](https://mlsanity.com/paper/2401.09670/) (2024)
- [KVQuant: Towards 10 Million Context Length LLM Inference with KV Cache Quantization](https://mlsanity.com/paper/2401.18079/) (2024)
- [Splitwise: Efficient generative LLM inference using phase splitting](https://mlsanity.com/paper/2311.18677/) (2023)
- [Gated Delta Networks: Improving Mamba2 with Delta Rule](https://mlsanity.com/paper/2412.06464/) (2024)
- [CacheBlend: Fast Large Language Model Serving for RAG with Cached Knowledge Fusion](https://mlsanity.com/paper/2405.16444/) (2024)
- [CacheGen: KV Cache Compression and Streaming for Fast Large Language Model Serving](https://mlsanity.com/paper/2310.07240/) (2023)
- [Mooncake: A KVCache-centric Disaggregated Architecture for LLM Serving](https://mlsanity.com/paper/2407.00079/) (2024)
- [Kimi Linear: An Expressive, Efficient Attention Architecture](https://mlsanity.com/paper/2510.26692/) (2025)
- [MiMo-V2-Flash Technical Report](https://mlsanity.com/paper/2601.02780/) (2026)
- [DynamoLLM: Designing LLM Inference Clusters for Performance and Energy Efficiency](https://mlsanity.com/paper/2408.00741/) (2024)
- [Nemotron 3 Nano: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning](https://mlsanity.com/paper/2512.20848/) (2025)
- [Lightning Attention-2: A Free Lunch for Handling Unlimited Sequence Lengths in Large Language Models](https://mlsanity.com/paper/2401.04658/) (2024)
- [Helix: Serving Large Language Models over Heterogeneous GPUs and Network via Max-Flow](https://mlsanity.com/paper/2406.01566/) (2024)
- [Step 3.5 Flash: Open Frontier-Level Intelligence with 11B Active Parameters](https://mlsanity.com/paper/2602.10604/) (2026)
- [LLM-PQ: Serving LLM on Heterogeneous Clusters with Phase-Aware Partition and Adaptive Quantization](https://mlsanity.com/paper/2403.01136/) (2024)
- [Hetis: Serving LLMs in Heterogeneous GPU Clusters with Fine-grained and Dynamic Parallelism](https://mlsanity.com/paper/2509.08309/) (2025)
- [From Prefix Cache to Fusion RAG Cache: Accelerating LLM Inference in Retrieval-Augmented Generation](https://mlsanity.com/paper/2601.12904/) (2026)
- [FREESH: Fair, Resource- and Energy-Efficient Scheduling for LLM Serving on Heterogeneous GPUs](https://mlsanity.com/paper/2511.00807/) (2025)

## Built on by (most cited first, 11 of 11)

- [KVServe: Service-Aware KV Cache Compression for Communication-Efficient Disaggregated LLM Serving](https://mlsanity.com/paper/2605.13734/) (2026)
- [MASPrism: Lightweight Failure Attribution for Multi-Agent Systems Using Prefill-Stage Signals](https://mlsanity.com/paper/2605.07509/) (2026)
- [SplitZip: Ultra Fast Lossless KV Compression for Disaggregated LLM Serving](https://mlsanity.com/paper/2605.01708/) (2026)
- [KVBuffer: IO-aware Serving for Linear Attention](https://mlsanity.com/paper/2605.19049/) (2026)
- [ObjectCache: Layerwise Object-Storage Retrieval for KV Cache Reuse](https://mlsanity.com/paper/2605.22850/) (2026)
- [PDD: Unleashing Economical and Flexible Heterogeneous LLM Inference via Cross-Datacenter Prefill-Decode Disaggregation](https://mlsanity.com/paper/2609.13161/) (2026)
- [PreFT: Prefill-only finetuning for efficient inference](https://mlsanity.com/paper/2605.14217/) (2026)
- [Lodestar: An Online-Learning LLM Inference Router](https://mlsanity.com/paper/2606.00946/) (2026)
- [TensorCast: The Missing Tensor Management Layer in Large Language Model Infrastructure](https://mlsanity.com/paper/2608.06007/) (2026)
- [Unified AI Gateway: A Framework for Joint Model Routing and KV Cache Management](https://mlsanity.com/paper/2609.06940/) (2026)
- [SARA: SLO-Aware Resource Allocation for Disaggregated Agentic LLM Services](https://mlsanity.com/paper/2609.26763/) (2026)

## Abstract (arXiv)

Prefill-decode (PD) disaggregation has become the standard architecture for large-scale LLM serving, but in practice its deployment boundary is still determined by KVCache transfer. In conventional dense-attention models, prefill generates huge KVCache traffics that keep prefill and decode tightly coupled within a single high-bandwidth network domain, limiting heterogeneous deployment and resource elasticity. Recent hybrid-attention architectures substantially reduce KVCache size, making cross-cluster KVCache transport increasingly plausible. However, smaller KVCache alone does not make heterogeneous cross-datacenter PD serving practical: real workloads remain bursty, request lengths are highly skewed, prefix caches are unevenly distributed, and inter-cluster bandwidth fluctuates. A naive design that fully externalizes prefill can therefore still suffer from congestion, unstable queueing, and poor utilization. We present Prefill-as-a-Service (PrfaaS), a cross-datacenter serving architecture that selectively offloads long-context prefill to standalone, compute-dense prefill clusters and transfers the resulting KVCache over commodity Ethernet to local PD clusters for decode. Rather than treating reduced KVCache as sufficient, PrfaaS combines model-side KV efficiency with system-side selective offloading, bandwidth-aware scheduling, and cache-aware request placement. This design removes the requirement that heterogeneous accelerators share the same low-latency RDMA fabric, enabling independent scaling of prefill and decode capacity across loosely coupled clusters. In a case study using an internal 1T-parameter hybrid model, a PrfaaS-augmented heterogeneous deployment achieves 54% higher serving throughput and 64% lower P90 TTFT than a homogeneous PD baseline, with approximately 15% throughput gain at equal cost, while consuming only modest cross-datacenter bandwidth.