Towards Load-Aware Prefill Deflection for Disaggregated LLM Serving
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.
VelServe is a scheduler designed to reduce decode jitters in large language model (LLM) inference, ensuring that latency service-level objectives (SLOs) for time to first token (TTFT) and time between tokens (TBT) are met. It calibrates the amount of prefill work each decode step performs based on the measured tail latency of that step, leaving headroom to prevent overruns and prioritizing shorter waiting prompts. Experiments on multiple attention families and workloads show that VelServe improves SLO goodput by 2.4–2.9× over the best baseline on Qwen3‑8B and DeepSeek‑V4‑Flash, and by at least 4.4× on DeepSeek‑V2‑Lite, while meeting SLOs for more than twice as many requests when prefill and decode are disaggregated across eight GPUs.
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:2607. 04206v1 Announce Type: cross Abstract: Diffusion language models (dLLMs) generate text by iteratively denoising a masked response and can commit multiple output positions per model invocation.
Diffusion language models (dLLMs) generate text by iteratively denoising a masked response and can commit multiple output positions per model invocation. Their bidirectional attention prevents exact autoregressive-style KV caching, since committing one position shifts the KV activations of all others.
arXiv:2608. 06557v1 Announce Type: cross Abstract: The reasoning and agentic capabilities of large language models have expanded the range of applications they support, from short interactive exchanges to long, compute-heavy requests.
arXiv:2610.18143v1 Announce Type: new Abstract: Prefix reuse improves LLM inference by sharing the KV cache of common prompt prefixes. This helps only if the scheduler can execute a matched branch in...
FluidPD is a prefill‑decode disaggregated LLM serving system that introduces in‑place elasticity to meet SLOs. It uses FluidToken to offload a bounded portion of prefill work to decode workers during short‑term imbalance, and FluidRole to reassign running workers between prefill and decode roles for sustained imbalance, all guided by lightweight pressure indices. In production Azure trace workloads, FluidPD improves overall SLO attainment over static SGLang by up to 94.6 percentage points without adding workers.
The paper introduces PACE, a per‑step admission controller that manages intra‑request parallelism in large language model serving. By admitting extra decoding branches only when the predicted branch externality fits within the batch’s current slack, PACE avoids the pitfalls of eager or fixed‑cap branch admission. Experiments on Qwen3‑32B and other large models show that PACE boosts goodput by up to 2.15× while maintaining 99% SLO attainment.
The paper investigates how prefix caching in large language model (LLM) agents can lead to persistent degradation when small bursts of additional requests occur. Even with identical base workloads, a brief surge can push a vLLM server into a degraded state—reducing hit ratios, throughput, and causing a backlog that grows for over an hour—while a server without the burst remains healthy. The study identifies that slower service keeps more sessions active, expanding the working set beyond the key‑value pool and triggering repeated recomputations, and demonstrates that simple mitigation strategies such as admission throttling, session suspension, and a budgeter that limits resident KV demand can prevent or recover from this collapse.
arXiv:2608.21719v1 Announce Type: cross Abstract: AI inference clusters are increasingly constrained by instantaneous power, not just energy: grid operators condition new capacity on demand response,...
arXiv:2607. 02630v1 Announce Type: cross Abstract: Hardware accelerators now sit on the critical path of online serving.
The paper introduces SafeStart, a technique for Mixture-of-Experts (MoE) language models that mitigates the safety degradation caused by active expert reduction during serving. By running all experts for the prompt and the first eight generated tokens before switching to a reduced set, SafeStart largely restores the model’s refusal to harmful requests while preserving most of the speedup achieved by expert reduction. Experiments on several MoE models show that SafeStart maintains high offline decode speedups (≈85% of the original) and limits the increase in harmful responses to within a few percentage points, even under jailbreak prompts.
The paper investigates how prefix caching in LLM agent serving can lead to persistent degradation when traffic bursts occur. It shows that even a small increase in arrivals can push a vLLM server into a low hit‑ratio, low‑throughput regime that continues to deteriorate for an extended period, while a server without prefix caching remains stable. The authors identify a workload‑balance condition that predicts this collapse and evaluate mitigation strategies such as throttling, session suspension, and a KV‑budgeting approach, finding the latter most effective.