Say what apps that change on a running host showed (017 stage 3, research 115)

Changing an app in place keeps a million items' app answering but for
1.12 ms (a host restart: 150 ms; stopping and starting it again from its
store, as first built, 150.4 ms), 171.5 ms when its records change shape
(332 ms). Another app's p99 stays 0.97 times as it was while one changes
every 0.2 s; a host of one app is as it was (TPC-C +0.7%, instructions
at most +0.02%); each module change leaves 19 KB. 030 holds the windows
of both builds, the isolation, the pairs, the counts and the memory.

1 files changed, 45 insertions, 0 deletions

diff --git a/research/experiments/030-app-lifecycle/pairs.py b/research/experiments/030-app-lifecycle/pairs.py
new file mode 100644
index 0000000..2ea4f48
--- /dev/null
+++ b/research/experiments/030-app-lifecycle/pairs.py
@@ -0,0 +1,45 @@
+#!/usr/bin/env python3
+"""The pairs of runs.sh pairs as a table (research 115): each round's TPC-C
+(transactions a second, CPU a transaction, the owner's core share) before
+and after, the medians and the bank queries' median times.
+
+pairs.py DATE   from DATE-tpcc-*.jsonl and DATE-queries-*.jsonl beside it
+"""
+import json
+import statistics
+import sys
+from pathlib import Path
+
+HERE = Path(__file__).resolve().parent
+day = sys.argv[1]
+
+
+def tpcc(version, round_):
+    for line in (HERE / f'{day}-tpcc-{version}-{round_}.jsonl').read_text().splitlines():
+        r = json.loads(line)
+        if r.get('step') == 'run' and r.get('system') == 'pure':
+            owner = max((share for name, share in r['core_shares'].items() if 'owner' in name), default=None) \
+                if isinstance(r['core_shares'], dict) else None
+            return r['transactions_per_second'], r['cpu_us_per_transaction'], owner
+
+
+def queries(version, round_):
+    return {r['query']: r['pure_median_ms'] for r in map(json.loads, (HERE / f'{day}-queries-{version}-{round_}.jsonl')
+                                                        .read_text().splitlines()) if 'query' in r}
+
+
+rounds = sorted({int(p.stem.rsplit('-', 1)[1]) for p in HERE.glob(f'{day}-tpcc-before-*.jsonl')})
+rows = {'before': [], 'in-place': []}
+for round_ in rounds:
+    b, a = tpcc('before', round_), tpcc('in-place', round_)
+    rows['before'].append(b), rows['in-place'].append(a)
+    print(json.dumps({'round': round_, 'tps': [b[0], a[0]], 'change': round((a[0] / b[0] - 1) * 100, 1),
+                      'cpu_us': [b[1], a[1]], 'owner_share': [b[2], a[2]]}))
+med = {v: statistics.median(t for t, _, _ in rows[v]) for v in rows}
+cpu = {v: statistics.median(c for _, c, _ in rows[v]) for v in rows}
+names = queries('before', rounds[0]).keys()
+q = {v: {n: statistics.median(queries(v, r)[n] for r in rounds) for n in names} for v in rows}
+print(json.dumps({'tps_median': [med['before'], med['in-place']], 'change': round((med['in-place'] / med['before'] - 1) * 100, 1),
+                  'cpu_us_median': [cpu['before'], cpu['in-place']],
+                  'queries_median_ms': {n: [q['before'][n], q['in-place'][n],
+                                            round((q['in-place'][n] / q['before'][n] - 1) * 100, 1)] for n in names}}))