Mooncake is a serving platform for Kimi, a leading LLM service by Moonshot AI, that uses a KVCache‑centric disaggregated architecture to separate prefill and decoding clusters. It leverages underutilized CPU, DRAM, and SSD resources in the GPU cluster to create a disaggregated KVCache and employs a scheduler that balances throughput and latency SLOs. To handle highly overloaded scenarios, Mooncake introduces a prediction‑based early rejection policy, achieving up to a 525% throughput increase in simulations and enabling Kimi to process 75% more requests under real workloads.
arXiv:2606. 24506v1 Announce Type: cross Abstract: Emerging LLM services increasingly host many sparse MoE models, yet most models receive sparse requests and remain cold.
arXiv:2508. 03611v3 Announce Type: replace-cross Abstract: This paper presents Astrolabe, a randomized prediction-guided scheduler for one-shot request dispatch in multi-instance large language model (LLM) serving.
arXiv:2606. 10493v1 Announce Type: cross Abstract: Local deployment of large Mixture-of-Experts (MoE) models falls short of the service quality achieved in cloud-scale environments, even under low-concurrency workloads.
SparKV is an adaptive framework for on-device large language model inference that intelligently decides whether to stream or compute each key-value (KV) cache chunk, overlapping communication and computation to cut latency. It dynamically adjusts offline-generated schedules at runtime to cope with changing wireless connectivity and edge resource availability. Experiments show SparKV can reduce Time-to-First-Token by 1.3×–5.1× and lower per-request energy consumption by 1.5×–3.3× with minimal impact on response quality.
arXiv:2609.16215v1 Announce Type: new Abstract: GPU high bandwidth memory is scarce and expensive, and KV caches consume much of it as chats, agent loops, and document question answering accumulate s...
The paper evaluates a learned request‑routing policy for disaggregated large‑language‑model serving, where compute‑heavy prefill and memory‑heavy decode stages run on separate GPU pools. Using a discrete‑event simulator and real NVIDIA A40 GPUs, the calibrated router—leveraging prompt length, predicted output length, KV‑cache pressure, and SLO class—outperforms round‑robin, least‑loaded, and length‑based heuristics, achieving the highest mean goodput (0.864) and lowest variance across three mixed, bursty arrival traces. Hardware calibration proves critical, providing a 4.5‑point goodput boost and roughly 40 % of the tail‑latency advantage, and the learned router can match round‑robin performance with one fewer GPU in certain scenarios.
arXiv:2607. 02043v1 Announce Type: cross Abstract: Disaggregated LLM serving runs prefill and decode on separate GPU pools to keep the two phases from interfering.
arXiv:2602. 24044v2 Announce Type: replace-cross Abstract: Large Language Model (LLM) adapters enable low-cost model specialization, but introduce complex caching and scheduling challenges in distributed serving systems where hundreds of adapters must be hosted concurrently.
arXiv:2606. 06302v1 Announce Type: new Abstract: Multi-turn Large Language Model (LLM) serving is critical for consistent user experiences, yet the linear growth of the Key-Value (KV) cache imposes significant pressure on GPU memory and bandwidth.
The paper presents a method for distributing large language model inference across multiple Intel AI PCs by splitting the model into pipeline shards, each pre‑compiled into an OpenVINO graph. Three key techniques—beam_idx Gather to enable GPU optimizations, speculative decoding on stateful models, and interleaved micro‑batching—allow a two‑node Llama 3.1 8B INT4 pipeline to serve two users at 1.79× the throughput of a single‑node model, while a four‑node deployment can run a 70B model that no single PC can hold. The authors provide code, benchmark logs, and reproduction scripts on GitHub.
arXiv:2511. 04791v2 Announce Type: replace Abstract: Modern LLM serving systems must sustain high throughput while meeting strict latency SLOs across two distinct inference phases: compute-intensive prefill and memory-bound decode phases.
arXiv:2607. 26566v1 Announce Type: cross Abstract: Text-to-image (T2I) workflows are increasingly deployed on serverless platforms because users often compose customized workflows and invoke them intermittently.