arXiv subjects

Jianjiang Li

Publications and source records attributed to Jianjiang Li.

3 recordsLinked to original sources

arXiv2026

Decoupling Readiness from Release for Tail-Aware Scheduling of Agentic LLM Workflows

Agentic LLM workflows consist of sequences of model turns interleaved with tool interactions, so their end-to-end completion time depends not only on inference speed but also on when ready turns are released. Most runtimes release each turn immediately upon readiness. Under contention, this eager release policy can accumulate released but unfinished work; once submitted, those turns can no longer be reordered by the workflow-level policy, increasing tail latency. We present a tail-risk-aware turn release scheduling method that jointly decides which ready turn to release next and how much released but unfinished work to maintain. The method uses a mean--Conditional Value-at-Risk (CVaR) objective to capture the evolving tail risk of unfinished workflows, incorporates online estimates of turn work when prioritizing ready turns, and adapts the released work budget to observed queue pressure. We evaluate the method using real agent execution traces from software engineering tasks across multiple LLMs and workflow arrival rates. The method performs comparably to eager release under light load and substantially reduces the P95 of workflow flow time under contention, achieving up to a $3.50\times$ speedup.

cs.AI↗

arXiv2026

FlowPrefill: Decoupling Preemption from Prefill Scheduling Granularity to Mitigate Head-of-Line Blocking in LLM Serving

The growing demand for large language models (LLMs) requires serving systems to handle many concurrent requests with diverse service level objectives (SLOs). This exacerbates head-of-line (HoL) blocking during the compute-intensive prefill phase, where long-running requests monopolize resources and delay higher-priority ones, leading to widespread time-to-first-token (TTFT) SLO violations. While chunked prefill enables interruptibility, it introduces an inherent trade-off between responsiveness and throughput: reducing chunk size improves response latency but degrades computational efficiency, whereas increasing chunk size maximizes throughput but exacerbates blocking. This necessitates an adaptive preemption mechanism. However, dynamically balancing execution granularity against scheduling overheads remains a key challenge. In this paper, we propose FlowPrefill, a TTFT-goodput-optimized serving system that resolves this conflict by decoupling preemption granularity from scheduling frequency. To achieve adaptive prefill scheduling, FlowPrefill introduces two key innovations: 1) Operator-Level Preemption, which leverages operator boundaries to enable fine-grained execution interruption without the efficiency loss associated with fixed small chunking; and 2) Event-Driven Scheduling, which triggers scheduling decisions only upon request arrival or completion events, thereby supporting efficient preemption responsiveness while minimizing control-plane overhead. Evaluation on real-world production traces shows that FlowPrefill improves maximum goodput by up to 5.6$\times$ compared to state-of-the-art systems while satisfying heterogeneous SLOs.

cs.DC↗

arXiv2023

Significant improvement of lossy compression rate and speed of HPC data using perceptron parallelized compression

The escalating surge in data generation presents formidable challenges to information technology, necessitating advancements in storage, retrieval, and utilization. With the proliferation of artificial intelligence and big data, the "Data Age 2025" report forecasts an exponential increase in global data production. The escalating data volumes raise concerns about efficient data processing. The paper addresses the predicament of achieving a lower compression ratio while maintaining or surpassing the compression performance of state-of-the-art techniques. This paper introduces a lossy compression framework grounded in the perceptron model for data prediction, striving for high compression quality. The contributions of this study encompass the introduction of positive and negative factors within the relative-to-absolute domain transformation algorithm, the utilization of a three-layer perceptron for improved predictive accuracy, and data selection rule modifications for parallelized compression within compression blocks. Comparative experiments with SZ2.1's PW_REL mode demonstrate a maximum compression ratio reduction of 17.78%. The article is structured as follows: the introduction highlights the data explosion challenge; related work delves into existing solutions; optimization of mapping algorithms in the relative and absolute domains is expounded in Section 3,the design of the new compression framework is detailed in Section 4,In Section 5 we describe the whole process and give pseudo-code, and in Section 6, our solution is evaluated. Finally, in Section 7, we provide an outlook for future work.

cs.DC↗