beebrain/verification/live_batch_memory.json at main · Blamp26/beebrain · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
{
  "status": "verified",
  "date": "2026-09-18",
  "platform": {
    "os": "Microsoft Windows 11 Pro 10.0.26200",
    "python": "3.12.10",
    "numpy": "2.4.6",
    "numba": "0.67.0",
    "llvmlite": "0.49.0",
    "psutil": "7.2.2",
    "connectome_neurons": 139255,
    "connectome_connections": 2698236,
    "connectome_sha256": "fbf8d440ca1207c7573e1acdd2366f9d0beb9b533c1710f21681264f81b1cc49"
  },
  "measurement_method": {
    "service": "uvicorn app:app --host 127.0.0.1 --port 8765 with BEEBRAIN_BATCH_CAPACITY=100 and BEEBRAIN_BATCH_JIT=1",
    "process_selection": "The newest Python process running the service was selected by PowerShell Get-Process; the Minecraft JVM was not included.",
    "rss_semantics": "Windows Process.WorkingSet64, reported as resident working-set/RSS-like memory in MiB (bytes divided by 2^20).",
    "private_memory_semantics": "Windows Process.PrivateMemorySize64, reported as private committed bytes in MiB (bytes divided by 2^20).",
    "cohort_sampling": "Twenty 500-ms samples per cohort after the first live production batch; the 100-bee peak sample used thirty 500-ms samples.",
    "sustained_sampling": "Sixty one-second samples during a 100-bee production run.",
    "pre_connectome_capture": {
      "captured": false,
      "reason": "The current service constructs Connectome and BatchRuntime eagerly at module import, before Uvicorn exposes a live service process state that can be sampled before connectome load."
    },
    "post_load_jit_capture": "Snapshots were taken after service startup; /health confirmed the full connectome and batch_jit=true. The first live production batch was treated as the final warm-up boundary because it touches the production active-lane path."
  },
  "baseline_and_warmup": {
    "post_connectome_jit_warmup_rss_mib": [201.61, 201.37, 201.53, 201.33, 201.75],
    "post_connectome_jit_warmup_private_mib": [1189.29, 1189.05, 1189.30, 1189.88, 1189.58],
    "first_live_batch_1_bee_rss_mib": {
      "min": 268.58,
      "mean": 268.64,
      "max": 268.72
    },
    "interpretation": "The first active production batch raises resident working set by about 67 MiB as previously untouched runtime/JIT pages become resident. Subsequent cohort values are the steady production measurements."
  },
  "cohorts": [
    {
      "bees": 1,
      "service_pid": 24936,
      "samples": 20,
      "rss_mib": {"min": 268.58, "mean": 268.64, "max": 268.72},
      "private_mib": {"min": 1189.24, "mean": 1189.29, "max": 1190.34}
    },
    {
      "bees": 10,
      "service_pid": 21744,
      "samples": 20,
      "rss_mib": {"min": 270.47, "mean": 270.50, "max": 270.53},
      "private_mib": {"min": 1190.89, "mean": 1190.89, "max": 1190.89}
    },
    {
      "bees": 30,
      "service_pid": 26524,
      "samples": 20,
      "rss_mib": {"min": 268.95, "mean": 268.98, "max": 269.01},
      "private_mib": {"min": 1189.36, "mean": 1189.36, "max": 1189.36}
    },
    {
      "bees": 64,
      "service_pid": 2892,
      "samples": 20,
      "rss_mib": {"min": 271.45, "mean": 271.52, "max": 271.55},
      "private_mib": {"min": 1192.41, "mean": 1192.41, "max": 1192.41}
    },
    {
      "bees": 100,
      "service_pid": 38904,
      "samples": 30,
      "rss_mib": {"min": 273.78, "mean": 273.90, "max": 273.98},
      "private_mib": {"min": 1194.38, "mean": 1194.38, "max": 1194.38}
    }
  ],
  "sustained_100_bee": {
    "duration_seconds": 60,
    "samples": 60,
    "rss_start_mib": 274.19,
    "rss_end_mib": 274.32,
    "rss_min_mib": 274.19,
    "rss_mean_mib": 274.25,
    "rss_peak_mib": 274.35,
    "private_start_mib": 1194.27,
    "private_end_mib": 1195.21,
    "private_peak_mib": 1195.28,
    "rss_growth_mib": 0.13,
    "private_growth_mib": 0.94
  },
  "churn": {
    "policy_retention_seconds": 60,
    "removed_cohort": 100,
    "owned_lanes_at_seconds_5_through_55": 100,
    "owned_lanes_at_seconds_60_through_75": 0,
    "rss_after_reap_mib": 274.42,
    "fresh_cohort": 100,
    "fresh_cohort_owned_lanes": 100,
    "fresh_cohort_rss_mib": 274.47,
    "fresh_cohort_private_mib": 1195.28,
    "fresh_cohort_batch_requests_observed": 1347,
    "fresh_cohort_batch_results_observed": 134700,
    "capacity_exceeded": false,
    "state_transferred_between_uuids": false,
    "bounded": true
  },
  "conclusion": {
    "steady_state_rss_mib_range": "268.58-273.98 across clean 1/10/30/64/100-bee cohort processes",
    "peak_rss_mib": 274.47,
    "100_bee_sustained_memory_bounded": true,
    "churn_memory_bounded": true,
    "performance_resource_milestone": "complete",
    "remaining_overall_project_status": "partially_complete"
  }
}