Home / Docs-Technical WhitePaper / 27-EFT.WP.Packets.Light v1.0
Chapter 10 — Queueing, Delay & Conservation (Bursts / Guard Bits)
One-sentence goal: Establish a unified convention for queueing and delay on links and nodes (OCS/OPS) and a guard-bit conservation rule, with computable equations for configured vs measured dual forms, accompanying contracts, and manifest-ready interfaces.
I. Scope & Objects
- Inputs
- Framing & timebase: frame_spec = { T_epoch, T_f, T_slot, T_sym, T_guard } (see Chapter 3), unified clocks tau_mono / ts.
- Traffic & burstiness: arrival parameters { λ_pkt, L_pkt } (packets) or { A, H } (circuit Erlang), token-bucket burst ( r, p, σ ).
- Nodes & resources: service rate C (bit/s or sym/s), number of ports P, FDL set { d_i } (OPS), bufferless/buffered policy, guard-bit policy.
- Path & switching: route and crossbar matrix (see Chapter 9), label label = ( lambda, k_sub, S_sel ) (see Chapter 5).
- Outputs
- Delay decomposition: lat_total = lat_prop + lat_ser + lat_sw + lat_q + lat_proc;
- Conservation & budgets: T_guard_budget, utilization, collision/drop rates;
- Dual-form gap: configured (model/simulation) vs measured (timestamps) discrepancy delta_form_queue;
- Manifest: manifest.packet.queue.* plus a contract report.
- Boundary
Physical propagation/compensation: Chapters 2 / 6; switching/routing: Chapter 9. This chapter focuses on queueing, delay, and guard-bit conservation.
II. Terms & Variables
- Rates & arrivals: λ_pkt [1/s], E[L] [bit], utilization ρ = λ_pkt * E[L] / C; circuit offered load A (Erlangs).
- Service & buffering: μ = C / E[L] (mean service rate), buffer capacity K [pkt] or FDL set { d_i }.
- Delay components: propagation lat_prop; serialization lat_ser = L / C; switching & reconfiguration lat_sw; queueing lat_q; processing lat_proc.
- Guard bits: T_guard ≥ 0, conservation budget T_guard_budget.
- Token bucket / burst: ( r, p, σ ); network calculus ( σ, ρ ).
- Two forms: configured (analysis/simulation/network calculus) vs measured (timestamp differences), gap delta_form_queue.
- Dimensions: unit(lat_*) = "[T]", unit(ρ) = "1", unit(T_guard) = "[T]".
III. Postulates P610-*
- P610-1 (Two forms in parallel): Any published delay must include both configured and measured forms, and record delta_form_queue.
- P610-2 (Hierarchical conservation): T_slot = N_sym*T_sym + N_gap*T_guard must hold; across nodes, guard-bit consumption must not drive downstream T_guard below T_guard_min (see Chapter 3).
- P610-3 (Explicit measures): Time / event / set–domain integrals or counts are explicit: ( ∫_{t∈W} • dt ), ( ∑_{pkt∈W} • ).
- P610-4 (Dimensional compliance): All published quantities pass check_dim( y − f(x) ); any dB↔linear conversion is annotated in the manifest.
- P610-5 (Fail-Closed): If conservation/thresholds fail or the dual-form gap exceeds limits, enforce rate-limiting / guard-bit enlargement / bypass or rollback, and persist the strategy card.
IV. Minimal Equations S610-*
- Mean queueing delay (stochastic models)
- M/M/1:
lat_q = ρ / ( μ (1 - ρ) ) = ( λ_pkt / μ ) / ( μ - λ_pkt );
total delay lat_total = lat_prop + E[L]/C + lat_sw + lat_q + lat_proc. - M/D/1:
lat_q = ( ρ^2 ) / ( 2 μ (1 - ρ) ). - Finite buffer M/M/1/K:
blocking P_block = ( (1-ρ) ρ^K ) / ( 1 - ρ^{K+1} ); effective delays computed conditional on K.
- M/M/1:
- OPS without buffer / with FDL (approximations)
- No buffer, no conversion: single-output contention
P_cont ≈ 1 - (1 - ρ / P_out)^{N_in}; drop rate ≈ P_cont. - With FDL { d_i }: schedulability P_sched = P(∃ d_i ≥ Δt_conflict), effective service μ_eff = μ • P_sched, ρ_eff = λ_pkt / μ_eff,
lat_q ≈ ρ_eff / ( μ_eff (1 - ρ_eff) ), bound lat_FDL ≤ max(d_i).
- No buffer, no conversion: single-output contention
- Network calculus (deterministic bounds)
- Arrival curve ( σ, ρ ) and service curve β(t) = R(t - T_0)^+:
- Delay bound D_max ≤ σ / (R - ρ) + T_0 (with R > ρ);
- Backlog bound B_max ≤ σ + ρ T_0.
- Guard-bit budget: T_guard_budget ≥ D_max + jitter_margin.
- Arrival curve ( σ, ρ ) and service curve β(t) = R(t - T_0)^+:
- Token-bucket shaping/policing
Token bucket ( r, p ): instantaneous arrivals bounded by A(t) ≤ r t + σ; shaper output satisfies ( σ_out, r ), shaping delay bound W_shaper ≤ σ / ( C - r ). - End-to-end delay composition
- lat_total = ∑_{links} ( lat_prop + E[L]/C + lat_sw + lat_q + lat_proc );
- 95th-percentile composition: lat_total,p95 ≈ ⊕_i lat_i,p95 (conservative summation or convolution under independence; manifest records the convention).
- Dual-form gap
delta_form_queue = | lat_total^{config} - lat_total^{meas} |; publication gate: delta_form_queue ≤ tol_queue.
V. Metrology Pipeline M60-10 (Ready → Model → Measure → Verify → Persist)
- Ready: freeze frame_spec and RefCond; gather node/port/FDL/rate and queueing policies; configure tol_queue / T_guard_min.
- Model: choose analytic forms (M/M/1, M/D/1, M/M/1/K, OPS contention/FDL) and/or network calculus ( σ, ρ; R, T_0 ); derive lat_total^{config} and T_guard_budget.
- Measure: use timestamps/counters to collect lat_sw / lat_q / lat_total, P_cont / P_block, and the actual T_guard consumption.
- Verify: compute delta_form_queue; validate T_guard ≥ 2σ_J + ΔCD + ΔPMD (see Chapter 3), P_cont / P_block, ρ and resource usage; trigger strategy cards (rate-limit / enlarge guard / reschedule / bypass).
- Persist:
manifest.packet.queue.* = { frame.hash, RefCond, model:{ type, params }, lat^{ config, meas }, delta_form_queue, ρ, P_cont / P_block, T_guard_budget, contracts.*, signature }.
VI. Contracts & Assertions C60-10x (Suggested Thresholds)
- C60-1001 (Dual-form gap): delta_form_queue_p95 ≤ tol_queue (typical tol_queue = 0.05 • lat_total^{config}).
- C60-1002 (Guard-bit conservation): T_guard ≥ 2σ_J + ΔCD + ΔPMD and T_guard ≥ T_guard_min.
- C60-1003 (Load & stability): ρ < 1 (or the calculus condition R > ρ); else rate-limit or scale out.
- C60-1004 (Contention / blocking): P_cont ≤ P_cont_max, P_block ≤ P_block_max.
- C60-1005 (Dimensional compliance): all lat_* / ρ / P_* / T_guard fields pass check_dim.
- C60-1006 (Freshness & coverage): measurement window and panel updates ≤ Δt_panel_max, coverage ≥ cov_min.
VII. Implementation Bindings I60-10* (interfaces, I/O, invariants)
- I60-101 queue_model(params:{ λ, E[L], C, K }) -> { lat_q, P_block }
- I60-102 ops_fdlsched(params:{ λ, μ, P_out, N_in, FDL, WC }) -> { P_cont, μ_eff, lat_q }
- I60-103 netcal_bound(arrival:{ σ, ρ }, service:{ R, T0 }) -> { D_max, B_max }
- I60-104 shaper_delay(token_bucket:{ r, σ }, C) -> { W_shaper }
- I60-105 compose_latency(links[]) -> { lat_total_{ mean, p95 } }
- I60-106 guard_budget(jitter, residuals:{ ΔCD, ΔPMD }) -> { T_guard_min }
- I60-107 measure_latency(ts_stream) -> { lat_prop, lat_ser, lat_sw, lat_q, lat_proc, lat_total }
- I60-108 compare_dual_queue(config, meas) -> delta_form_queue
- I60-109 emit_queue_manifest(results, policy) -> manifest.packet.queue
Invariants: two_forms_present = true; check_dim(*) passes; frames and routes are traceable; rate-limit / guard-bit expansion actions persisted.
VIII. Cross-References
- Framing/alignment & guard bits: Chapter 3;
- Switching & routing: Chapter 9;
- Arrival-time harmonization: Chapter 8;
- Physical compensation & residuals: Chapter 6 (for ΔCD / ΔPMD in the conservation term);
- Metrology & dashboards: Chapter 11 (delay / load / contention panel fields).
IX. Quality & Risk Control
- SLI / SLO: delta_form_queue_p95, lat_total_p95, ρ_p95, P_cont / P_block, T_guard_violation_rate, throughput_stability.
- Fallback strategies:
- ρ ≥ 1 or rising P_cont: rate-limit / scale out / reschedule;
- Rising delta_form_queue: back-write model parameters or prefer measured form;
- Guard shortfall: increase T_guard, reduce throughput, or defer scheduling;
- SLA delay exceedance: priority / reserved slices or bypass.
- Audit: models/parameters and measurement evidence, threshold changes, strategy-card execution logs, and the manifest signature chain with replay consistency.
Summary
- Centering on queueing/delay and guard-bit conservation, this chapter provides unified equations and publication rules across M/M/1, FDL, and network-calculus formulations.
- Through P610 / S610 / M60-10 / C60-10x / I60-10* and the manifest.packet.queue.* schema, queueing and delay become measurable, auditable, and rollback-ready.
Copyright & License (CC BY 4.0)
Copyright: Unless otherwise noted, the copyright of “Energy Filament Theory” (text, charts, illustrations, symbols, and formulas) belongs to the author “Guanglin Tu”.
License: This work is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0). You may copy, redistribute, excerpt, adapt, and share for commercial or non‑commercial purposes with proper attribution.
Suggested attribution: Author: “Guanglin Tu”; Work: “Energy Filament Theory”; Source: energyfilament.org; License: CC BY 4.0.
First published: 2025-11-11|Current version:v5.1
License link:https://creativecommons.org/licenses/by/4.0/