MB5 PD reuse-centric ablation: tooling, data, Fig 1-3

Three-axis controlled ablation of PD-colo vs PD-disagg on synthetic regular
traces (closed-loop, controlled reuse via REPLAY_NO_REALIZED_PREFIX) on the
clean stack (e13391e gated off).

  Axis 1 (Fig 1) -- reuse 6%->94% at N=8, in8192/out256
  Axis 2 (Fig 2) -- shape in2048/out2048 -> in32768/out64 at N=8, reuse~70%
  Axis 3 (Fig 3) -- concurrency N=8/16/32/64 at reuse~71%, in8192/out256

Findings:
  * APC parity colo=PD at every reuse (5.5/22/44/66/77/82%) -- contamination
    fix validated.
  * PD edge erodes 1.57x->1.10x with reuse; prefill GPUs strand 26%->9%.
  * Shape: PD-best peaks mid-sweep (1.34x at in8192/out512); wrong PD ratio
    catastrophic at prefill extreme (in32768/out64 pd2 = 378/400, p99 432s).
  * Concurrency: PD wins N<=32 (1.23-1.29x), TIPS at N=64 -- pd2/pd4
    crater (APC 71%->1.4%, TPS -30%) while colo scales cleanly.

Infrastructure:
  * replayer: --max-inflight-sessions, --inter-turn-think, --no-realized-prefix
    (env-defaulted via REPLAY_MAX_INFLIGHT, REPLAY_INTER_TURN_THINK_S,
    REPLAY_NO_REALIZED_PREFIX).
  * mb5_run.sh: writes bench_config.json + gpu_util.csv + run_window.json +
    instance_apc.txt + metrics.jsonl for bench_report/fig_agg ingest.
  * fig_agg.py: per-arm GPU role split + producer-side APC; --json mode.
  * gpu_util_report.py: companion per-GPU util report from gpu_util.csv.
  * partial_summary.py: stats from in-flight replay_metrics.jsonl
    (works before metrics.summary.json exists).

Data: analysis/mb5_pd_ablation/fig{1,2,3}.json (24 + 20 + 16 rows).
Figures: figs/mb5_pd_ablation/fig{1_reuse,2_shape,3_concurrency}_axis.png.
This commit is contained in:
2026-05-31 20:14:46 +08:00
parent a2111b6e18
commit fafc44da79
12 changed files with 389 additions and 9 deletions

View File

@@ -0,0 +1,140 @@
"""Aggregate a set of MB5 run dirs into one comparison table.
Pulls the three core metrics the analysis cares about, per run:
- E2E latency (from replay_metrics.summary.json: latency_stats_s)
- TPS (output tokens / wall_clock_s)
- GPU util by workers (gpu_util.csv over run_window, split prefill/decode by role)
plus honest reuse (producer-side APC from instance_apc.txt) and TTFT/TPOT for logs.
Arm + GPU role split + producer APC ports are inferred from the dir name:
*_colo_* -> 8 kv_both ; apc ports 8000-8007 (all keep prefix)
*_pd6_* -> 6P+2D P0-5/D6-7 ; apc 8000-8005
*_pd_* -> 4P+4D P0-3/D4-7 ; apc 8000-8003 (note: "pd" not "pd4")
*_pd2_* -> 2P+6D P0-1/D2-7 ; apc 8000-8001
Usage: fig_agg.py <run_dir> [<run_dir> ...]
"""
from __future__ import annotations
import csv
import json
import re
import statistics
import sys
from pathlib import Path
def arm_of(name: str):
# New driver naming (run_conc.sh / run_reuse_fixed.sh): "..._<CONFIG>_rep<r>".
if "8C-proxy" in name:
return "colo", list(range(8)), [], list(range(8000, 8008))
if "6P+2D" in name:
return "6P+2D", [0, 1, 2, 3, 4, 5], [6, 7], list(range(8000, 8006))
if "2P+6D" in name:
return "2P+6D", [0, 1], [2, 3, 4, 5, 6, 7], list(range(8000, 8002))
if "4P+4D" in name:
return "4P+4D", [0, 1, 2, 3], [4, 5, 6, 7], list(range(8000, 8004))
# Legacy naming (original May-30 corrected runs).
if "_colo_" in name or name.endswith("_colo"):
return "colo", list(range(8)), [], list(range(8000, 8008))
if "_pd6_" in name:
return "6P+2D", [0, 1, 2, 3, 4, 5], [6, 7], list(range(8000, 8006))
if "_pd2_" in name:
return "2P+6D", [0, 1], [2, 3, 4, 5, 6, 7], list(range(8000, 8002))
if "_pd4_" in name or "_pd_" in name:
return "4P+4D", [0, 1, 2, 3], [4, 5, 6, 7], list(range(8000, 8004))
return "?", list(range(8)), [], list(range(8000, 8008))
def util_split(run: Path, pgpus, dgpus):
win = {}
wp = run / "run_window.json"
if wp.exists():
win = json.load(open(wp))
t0, t1 = win.get("t_start_unix"), win.get("t_end_unix")
csvp = run / "gpu_util.csv"
if not csvp.exists():
return None, None
by = {}
for row in csv.DictReader(open(csvp)):
try:
ts = float(row["timestamp"]); g = int(row["gpu"]); u = float(row["util_pct"])
except (ValueError, KeyError):
continue
if t0 and not (t0 <= ts <= t1):
continue
by.setdefault(g, []).append(u)
pm = [v for g in pgpus for v in by.get(g, [])]
dm = [v for g in dgpus for v in by.get(g, [])]
return (statistics.fmean(pm) if pm else None,
statistics.fmean(dm) if dm else None)
def apc(run: Path, ports):
f = run / "instance_apc.txt"
if not f.exists():
return None
q = h = 0
for line in open(f):
m = dict(re.findall(r"(\w+)=(\S+)", line))
try:
p = int(m.get("port", -1))
except ValueError:
continue
if p in ports:
q += float(m.get("queries", 0)); h += float(m.get("hits", 0))
return (h / q) if q else None
def main():
args = sys.argv[1:]
as_json = False
if "--json" in args:
as_json = True
args = [a for a in args if a != "--json"]
rows = []
for d in args:
run = Path(d)
sp = run / "replay_metrics.summary.json"
if not sp.exists():
continue
s = json.load(open(sp))
arm, pg, dg, ports = arm_of(run.name)
lat = s.get("latency_stats_s", {})
ttft = s.get("ttft_stats_s", {})
tpot = s.get("tpot_stats_s", {})
wall = s.get("wall_clock_s") or 1.0
out = s.get("actual_output_tokens_stats", {})
n = s.get("success_count", 0); req = s.get("request_count", 0)
tot_out = out.get("count", 0) * out.get("mean", 0)
tps = tot_out / wall
pu, du = util_split(run, pg, dg)
a = apc(run, ports)
rows.append({
"name": run.name, "arm": arm, "n": n, "req": req,
"e2e_p50": lat.get("p50"), "e2e_p90": lat.get("p90"), "e2e_p99": lat.get("p99"),
"e2e_mean": lat.get("mean"),
"ttft_p90": ttft.get("p90"), "tpot_p99": tpot.get("p99"),
"tps": tps, "wall": wall, "pu": pu, "du": du, "apc": a,
})
if as_json:
print(json.dumps(rows))
return
def f(x, w=7, p=1):
return f"{x:>{w}.{p}f}" if isinstance(x, (int, float)) else f"{'-':>{w}}"
hdr = (f"{'run':<34}{'arm':>7}{'ok/req':>9}{'E2Ep50':>8}{'E2Ep90':>8}{'E2Ep99':>8}"
f"{'TPS':>8}{'Putil':>7}{'Dutil':>7}{'APC%':>7}{'TTFTp90':>9}{'TPOTp99ms':>10}")
print(hdr); print("-" * len(hdr))
for r in sorted(rows, key=lambda r: r["name"]):
print(f"{r['name']:<34}{r['arm']:>7}{str(r['n'])+'/'+str(r['req']):>9}"
f"{f(r['e2e_p50'])}{f(r['e2e_p90'])}{f(r['e2e_p99'])}"
f"{f(r['tps'],8,1)}{f(r['pu'])}{f(r['du'])}"
f"{f((r['apc'] or 0)*100)}{f(r['ttft_p90'],9,2)}"
f"{f((r['tpot_p99'] or 0)*1000,10,1)}")
if __name__ == "__main__":
main()

View File

@@ -0,0 +1,71 @@
"""Per-GPU utilization report from gpu_util.csv (companion to bench_report.py).
bench_report's per-worker GPU util needs request routing (breakdown.json), which
the MB5 proxy doesn't log. But worker == GPU by index, and the prefill/decode role
split is fixed by config, so per-GPU util from gpu_util.csv directly answers
"GPU utils by workers" — and for PD it exposes the key signal: are the prefill-side
GPUs saturated while the decode-side idles (or vice versa, or stalled at ~0)?
Usage:
gpu_util_report.py <run_dir> [--prefill-gpus 0,1,2,3 --decode-gpus 4,5,6,7]
"""
from __future__ import annotations
import argparse
import csv
import json
import statistics
from pathlib import Path
def pct(xs, p):
xs = sorted(xs)
return xs[max(0, min(len(xs) - 1, int(round(p / 100 * (len(xs) - 1)))))] if xs else None
def main():
ap = argparse.ArgumentParser()
ap.add_argument("run_dir", type=Path)
ap.add_argument("--prefill-gpus", default="")
ap.add_argument("--decode-gpus", default="")
a = ap.parse_args()
win = {}
wp = a.run_dir / "run_window.json"
if wp.exists():
win = json.load(open(wp))
t0, t1 = win.get("t_start_unix"), win.get("t_end_unix")
csvp = a.run_dir / "gpu_util.csv"
if not csvp.exists():
print(f"{a.run_dir.name}: gpu_util.csv absent"); return
by_gpu = {}
for row in csv.DictReader(open(csvp)):
try:
ts = float(row["timestamp"]); g = int(row["gpu"]); u = float(row["util_pct"]); m = float(row["mem_used_mb"])
except (ValueError, KeyError):
continue
if t0 and not (t0 <= ts <= t1):
continue
by_gpu.setdefault(g, {"u": [], "m": []})
by_gpu[g]["u"].append(u); by_gpu[g]["m"].append(m)
print(f"=== {a.run_dir.name}: per-GPU util over replay window ({sum(len(d['u']) for d in by_gpu.values())} samples) ===")
print(f"{'gpu':>4}{'util_mean':>11}{'util_p90':>10}{'util_max':>10}{'mem_max_GB':>12}")
for g in sorted(by_gpu):
u, m = by_gpu[g]["u"], by_gpu[g]["m"]
print(f"{g:>4}{statistics.fmean(u):>11.1f}{pct(u,90):>10.1f}{max(u):>10.1f}{max(m)/1024:>12.1f}")
def agg(gpus, label):
gpus = [int(x) for x in gpus.split(",") if x != ""]
us = [v for g in gpus for v in by_gpu.get(g, {}).get("u", [])]
if us:
print(f" {label:<14} gpus={gpus} util mean={statistics.fmean(us):.1f}% p90={pct(us,90):.1f}% max={max(us):.1f}%")
if a.prefill_gpus:
agg(a.prefill_gpus, "prefill-side")
if a.decode_gpus:
agg(a.decode_gpus, "decode-side")
if __name__ == "__main__":
main()

View File

@@ -69,6 +69,13 @@ run_one() {
source "${VENV}/bin/activate"
local replay_out="${rundir}/replay_metrics.jsonl"
mkdir -p "$(dirname "${replay_out}")"
# bench_report.py inputs: worker->gpu map (worker i == gpu i for every config;
# for PD, workers 0-3 are producers on gpu0-3, 4-7 consumers on gpu4-7).
printf '{"base_port":8000,"n_instances":8,"gpu_indices":[0,1,2,3,4,5,6,7]}\n' \
> "${rundir}/bench_config.json"
# per-GPU utilization timeseries over the replay window (2s sampling)
bash "${SCRIPT_DIR}/gpu_monitor.sh" "${rundir}/gpu_util.csv" 2 >/dev/null 2>&1 &
local GPU_MON=$!
local t0
t0=$(date +%s.%N)
if ! PYTHONPATH="${FRESH_ROOT}" python -m replayer \
@@ -82,6 +89,7 @@ run_one() {
t1=$(date +%s.%N)
local wall=$(python -c "print(${t1} - ${t0})")
echo "[mb5-run] REPLAY FAILED after ${wall} s; see ${OUT_ROOT}/${config}_rep${rep}_replay.log"
kill "${GPU_MON}" 2>/dev/null || true
bash "${LAUNCH}" stop > /dev/null 2>&1 || true
return 1
fi
@@ -91,6 +99,9 @@ run_one() {
wall_clock_s=$(python -c "print(${t1} - ${t0})")
echo "[mb5-run] replay done in ${wall_clock_s}s"
echo "${wall_clock_s}" > "${rundir}/wall_clock_s.txt"
kill "${GPU_MON}" 2>/dev/null || true
printf '{"t_start_unix":%s,"t_end_unix":%s}\n' "${t0}" "${t1}" > "${rundir}/run_window.json"
cp -f "${replay_out}" "${rundir}/metrics.jsonl" # bench_report.py expects metrics.jsonl
# Per-instance prefix-cache counters, scraped from each backend BEFORE
# teardown. For PD this is the only honest reuse signal: producer ports

View File

@@ -0,0 +1,98 @@
"""Compute a per-run summary directly from replay_metrics.jsonl (for partial / in-flight runs).
Used when the replayer hasn't completed (so replay_metrics.summary.json doesn't exist
yet) but enough records have streamed to disk to read out the per-arm result.
Also accepts a finished run's directory and prints the same one-line summary for
apples-to-apples comparison.
"""
from __future__ import annotations
import json
import re
import statistics
import sys
from pathlib import Path
def stats(xs):
xs = sorted(xs)
n = len(xs)
if n == 0:
return None
return {
"n": n,
"mean": statistics.fmean(xs),
"p50": xs[n // 2],
"p90": xs[int(0.9 * (n - 1))],
"p99": xs[int(0.99 * (n - 1))],
}
def apc(run: Path, producer_ports):
f = run / "instance_apc.txt"
if not f.exists():
return None
q = h = 0.0
for line in open(f):
m = dict(re.findall(r"(\w+)=(\S+)", line))
try:
p = int(m.get("port", -1))
except ValueError:
continue
if p in producer_ports:
q += float(m.get("queries", 0))
h += float(m.get("hits", 0))
return (h / q) if q else None
def main():
for d in sys.argv[1:]:
run = Path(d)
# prefer the live replay_metrics.jsonl (so partials work); fall back to metrics.jsonl
for fn in ("replay_metrics.partial.jsonl", "replay_metrics.jsonl", "metrics.jsonl"):
p = run / fn
if p.exists():
rec_path = p
break
else:
print(f"{run.name}: no records"); continue
recs = [json.loads(l) for l in open(rec_path)]
oks = [r for r in recs if r.get("error") is None]
lat = stats([r["latency_s"] for r in oks if "latency_s" in r])
ttft = stats([r["ttft_s"] for r in oks if "ttft_s" in r])
tpot = stats([r["tpot_s"] for r in oks if "tpot_s" in r])
out = sum(r.get("actual_output_tokens", r.get("output_length", 0)) for r in oks)
ts = [r["t_dispatch_unix"] for r in oks if "t_dispatch_unix" in r]
tf = [r["t_finish_unix"] for r in oks if "t_finish_unix" in r]
span = max(tf) - min(ts) if ts and tf else 0
tps = out / span if span else 0
# producer ports by arm tag in dirname
n = run.name
if "_colo_" in n:
ports = list(range(8000, 8008))
elif "_pd6_" in n:
ports = list(range(8000, 8006))
elif "_pd2_" in n:
ports = list(range(8000, 8002))
else:
ports = list(range(8000, 8004))
a = apc(run, ports)
print(f"{run.name}")
print(f" n_ok={len(oks)}/{len(recs)}"
+ (f" (target=1214 -> {len(oks)*100/1214:.1f}%)" if len(recs) < 1214 else ""))
if lat:
print(f" E2E mean={lat['mean']:.2f} p50={lat['p50']:.2f} p90={lat['p90']:.2f} p99={lat['p99']:.2f}")
if ttft:
print(f" TTFT mean={ttft['mean']:.2f} p50={ttft['p50']:.2f} p90={ttft['p90']:.2f} p99={ttft['p99']:.2f}")
if tpot:
print(f" TPOT mean={tpot['mean']*1000:.1f}ms p90={tpot['p90']*1000:.1f}ms p99={tpot['p99']*1000:.1f}ms")
print(f" output_tokens={out:.0f} span={span:.0f}s TPS={tps:.0f}")
if a is not None:
print(f" producer APC={a*100:.1f}%")
if __name__ == "__main__":
main()