Chronos — Prefill Decode Fabric

\documentclass[conference]{IEEEtran}
\IEEEoverridecommandlockouts
\usepackage{cite}
\usepackage{amsmath,amssymb,amsfonts}
\usepackage{algorithmic}
\usepackage{graphicx}
\usepackage{textcomp}
\usepackage{xcolor}
\usepackage{hyperref}
\usepackage{booktabs}
\usepackage{array}
\usepackage{tabularx}
\newcolumntype{Y}{>{\raggedright\arraybackslash}X}
\hypersetup{hidelinks}
\begin{document}

\title{CHRONOS Prefill-Decode Fabric: An Evidence-Aware Architecture, Prototype Evaluation, and Integration Roadmap}
\author{\IEEEauthorblockN{Aaron Singh}
\IEEEauthorblockA{\textit{Department of Electrical and Computer Engineering} \\
\textit{San Francisco State University}}}
\maketitle

\begin{abstract}
This paper presents CHRONOS Prefill-Decode Fabric, a project in a twenty-project AI infrastructure portfolio. The current repository is an active prototype with implemented behavior, automated correctness tests, and explicit boundaries around unverified hardware or production claims. We describe the system model, current implementation, goals, and evaluation method, then propose five integrations that advance the project toward reproducible system-level validation. A local audit on August 14, 2026 executed 3 project tests successfully. No accelerator, deployment, or performance conclusion is inferred unless a corresponding committed benchmark or profiler artifact exists.
\end{abstract}

\begin{IEEEkeywords}
CHRONOS, AI infrastructure, reproducibility, prototype validation, systems evaluation, hardware--software co-design
\end{IEEEkeywords}

\section{Introduction}
CHRONOS Prefill-Decode Fabric (P09) addresses a bounded problem within the portfolio's track-03-distributed-infrastructure track. Its declared status is \emph{Active Prototype} \cite{metadata}. The project validates its central mechanism before adding device-specific acceleration, production orchestration, or real-world hardware. This ordering matters because optimization without a trusted reference can make incorrect behavior appear successful.

The project goals are to establish deterministic behavior, encode correctness as tests, create machine-readable evidence, identify limiting resources through measurement, and integrate results with adjacent portfolio systems without losing provenance. The repository contains one paper named exactly after its singular folder: \texttt{P09-chronos-prefill-decode-fabric.tex}.

\section{Technical Context}
Distributed inference depends on scheduling, capacity, communication, reliability, and observability. Later multi-node experiments must record topology, synchronization costs, retries, and tail latency. The portfolio benchmark standard requires environment, workload, method, metric, artifact, reproduction, and limitation fields; unknown values remain explicitly unmeasured \cite{benchmark}.

\section{System Model and Architecture}
The prototype is organized around the following domain model:
\begin{equation}
T_i=W_i^{pre}+S_i^{pre}+W_i^{dec}+S_i^{dec}
\end{equation}
The equation is a design and test abstraction rather than a claimed empirical law. It supports invariants and expected-value checks while later implementations replace synthetic inputs with representative workloads or devices.

The software architecture contains an input/configuration layer, a deterministic core, validation and evidence output, and a local visualization. The principal inspected source artifacts are \texttt{python/scheduler.py}. Unsupported real-world conditions are surfaced as limitations rather than silently simulated.

\section{Detailed Script Operation and Rationale}
The scheduler models separate prefill and decode service stages. It advances queue timing, guarantees that each stage does not overlap itself, ensures decode follows prefill, and emits machine-readable timing data for later disaggregated-serving experiments.

The execution path is:
\begin{enumerate}
    \item Create deterministic requests with prefill and decode work.
    \item Queue and service prefill without overlapping the prefill worker.
    \item Release completed prefill work into the decode queue.
    \item Service decode without overlap and enforce stage ordering.
    \item Emit per-request timing for future transfer-cost and parallel-worker models.
\end{enumerate}

\begin{table*}[t]
\caption{Implementation artifacts and why they exist}
\label{tab:p09-implementation}
\centering
\small
\begin{tabularx}{\textwidth}{p{0.24\textwidth}YY}
\toprule
\textbf{Artifact} & \textbf{Observed responsibility} & \textbf{Engineering rationale} \\
\midrule
python/scheduler.py & Job, simulate, sample, main & Implements the inspectable, unit-tested project core. \\
scripts/reproduce.sh & Fixed test and demonstration entry point & Gives another developer one command for local reproduction. \\
PROJECT.yaml and ANALYSIS.md & Status, completed work, planned work, and claim boundaries & Separates declared intent from evidence-backed implementation. \\
streamlit\_app.py & Local evidence and status visualization & Makes outputs inspectable without upgrading simulation into a hardware claim. \\
\bottomrule
\end{tabularx}
\end{table*}

\section{Implemented Prototype}
The metadata and source audit found these completed features \cite{analysis}:
\begin{itemize}
    \item Python two-stage scheduler
    \item Prefill/decode queue timing
    \item Unit tests and JSONL output
\end{itemize}

On August 14, 2026, \texttt{python3 -m unittest discover -s tests -p 'test\_*.py'} completed successfully with 3 tests. The inspected test artifacts are \texttt{tests/test\_scheduler.py}. This is evidence of local correctness for encoded cases, not production scale or hardware performance.

\section{Testing Methodology and Observed Results}
Testing uses Python's standard \texttt{unittest} discovery and exercises the public behavior of the reference implementation. The audit reran the suite from the project folder with \texttt{python3 -m unittest discover -s tests -p 'test\_*.py'}. All 3 discovered tests passed. The result establishes correctness only for the encoded local cases; it does not establish accelerator correctness, real-device behavior, production reliability, or benchmark completion.

\begin{table*}[t]
\caption{Audited test matrix}
\label{tab:p09-tests}
\centering
\scriptsize
\begin{tabularx}{\textwidth}{p{0.37\textwidth}Yp{0.21\textwidth}}
\toprule
\textbf{Test artifact and case} & \textbf{Behavior being checked} & \textbf{Observed result} \\
\midrule
tests/test\_scheduler.py:test\_stages\_do\_not\_overlap\_themselves & Stages do not overlap themselves. & Pass (local, 2026-08-14) \\
tests/test\_scheduler.py:test\_decode\_follows\_prefill & Decode follows prefill. & Pass (local, 2026-08-14) \\
tests/test\_scheduler.py:test\_bad\_rate & Bad rate. & Pass (local, 2026-08-14) \\
\bottomrule
\end{tabularx}
\end{table*}

No numerical performance result is promoted by this test run. Where scripts emit JSON or JSONL, those outputs remain raw or simulation-specific until a reviewed summary includes hardware, software, workload, warm-up, repetition, correctness threshold, Git revision, and limitations.

\begin{table*}[t]
\caption{Declared status versus audited evidence}
\label{tab:p09-audit}
\centering
\small
\begin{tabularx}{\textwidth}{p{0.20\textwidth}Yp{0.25\textwidth}}
\toprule
\textbf{Audit field} & \textbf{Finding} & \textbf{Evidence source} \\
\midrule
Declared status & Active Prototype & PROJECT.yaml \\
Evidence-backed status & Active local prototype; 3 tests passed & Source plus local unittest run \\
Accepted measured results & None recorded in measured\_results & PROJECT.yaml \\
Mismatch / claim boundary & Simulation does not prove serving performance & PROJECT.yaml and ANALYSIS.md \\
Next proof required & Parallel workers; Network transfer cost & Planned features \\
\bottomrule
\end{tabularx}
\end{table*}

\section{Claim Boundaries and Risks}
The project records these unverified or excluded claims:
\begin{itemize}
    \item Simulation does not prove serving performance
\end{itemize}
The main risk is confusing synthetic or modeled behavior with deployed-system behavior. Other risks include incomplete workloads, platform-dependent timing, missing failure injection, and interfaces not yet exercised across device boundaries. Performance claims require a reviewed record meeting the portfolio standard.

\section{Evaluation Plan}
Evaluation proceeds through correctness tests, deterministic reproduction with Git and environment metadata, repeated benchmarks reporting latency/throughput/memory/error metrics, and a named profiler capture tied to exact hardware and source revision. Success requires reference equivalence within a documented tolerance, preservation of safety and resource invariants, clear failures, and evidence reproducible from a clean environment.

\section{Goals, Milestones, and Success Criteria}
The project goals are staged so that correctness precedes performance and integration. A goal is complete only when its proof artifact is committed or otherwise reviewable; prose or a simulated number alone is insufficient.

\begin{table*}[t]
\caption{Project goals and required proof}
\label{tab:p09-goals}
\centering
\small
\begin{tabularx}{\textwidth}{p{0.06\textwidth}YY}
\toprule
\textbf{ID} & \textbf{Goal} & \textbf{Completion evidence} \\
\midrule
G1 & Parallel workers & Passing tests and a reviewed source artifact \\
G2 & Network transfer cost & Machine-readable result with reproduction metadata \\
G3 & vLLM adapter & Reference-equivalence or domain-correctness report \\
G4 & Model transfer and parallel workers & Named profiler, deployment, or integration artifact \\
G5 & Validate a disaggregated serving adapter & Dashboard/report link preserving provenance and limitations \\
\bottomrule
\end{tabularx}
\end{table*}

\section{Future Work and Integrations}
The five project-specific next steps are:
\begin{enumerate}
    \item Parallel workers
    \item Network transfer cost
    \item vLLM adapter
    \item Add KV-cache transfer, backpressure, and recovery state machines.
    \item Deploy the fabric on P10's two-node testbed.
\end{enumerate}
The early items complete declared evidence; the later items connect downstream portfolio consumers. Each integration should add tests and a reviewable artifact such as JSONL evidence, a report, profiler capture, deployment manifest, trace, or labeled data set.

\section{Conclusion}
CHRONOS Prefill-Decode Fabric is an evidence-aware active prototype: its implemented behavior and tests are real, while unbuilt hardware, deployment, and performance goals remain labeled. Completing the five integrations in dependency order will advance it from a learning artifact toward a credible portfolio component.

\begin{thebibliography}{00}
\bibitem{metadata} Aaron Singh, ``CHRONOS Prefill-Decode Fabric PROJECT.yaml,'' local portfolio repository, updated 2026-08-13.
\bibitem{analysis} Aaron Singh, ``CHRONOS Prefill-Decode Fabric: README, ANALYSIS, source, and test artifacts,'' local portfolio repository, accessed Aug. 14, 2026.
\bibitem{benchmark} Aaron Singh, ``AI Infrastructure Portfolio Benchmark Standard,'' local portfolio repository, accessed Aug. 14, 2026.
\end{thebibliography}
\end{document}