Files
anonymous_github/backend/reports/zip-memory-improvement.json
T

95 lines
3.1 KiB
JSON

{
"fixture": {
"repository": "tdurieux/anonymous_github",
"commit": "b09ef0e5b0027756ad86eb3353a5ff6416016ade",
"originalFiles": 1017,
"originalBytes": 28687144,
"terms": [
"tdurieux"
],
"addedFixture": "1000 copies of snapshot gulpfile.js in benchmark-1000"
},
"output": {
"status": 200,
"sha256": "72d41a3fe5dcbf91eb33c1682cd4db5ccb220160c35e58e310ead96ac39ac600",
"entries": 2017
},
"method": "One fresh process per binary, one initial and one repeated archive download, sequential loopback runs. Peak RSS sampled every 20 ms during repeated request; includes worker processes, excludes MongoDB, Redis and benchmark client. No profiler attached. Filesystem caches were not dropped. Timing is indicative, not a multi-run statistical estimate.",
"before": {
"name": "repository ZIP",
"cold": {
"ms": 40264.038648,
"firstBodyMs": 177.61004400000093,
"status": 200,
"bytes": 13045815
},
"output": {
"status": 200,
"sha256": "72d41a3fe5dcbf91eb33c1682cd4db5ccb220160c35e58e310ead96ac39ac600",
"entries": 2017
},
"warm": [
{
"concurrency": 1,
"requests": 1,
"p50Ms": 39293.266413000005,
"p95Ms": 39293.266413000005,
"p50FirstBodyMs": 148.60325800000282,
"requestsPerSecond": 0.025418685391377722,
"cpuMs": 38020,
"peakProcessTreeRssBytes": 696705024,
"bytes": 13045799
}
]
},
"after": {
"name": "repository ZIP",
"cold": {
"ms": 36796.176851,
"firstBodyMs": 55.14695500000016,
"status": 200,
"bytes": 13045739
},
"output": {
"status": 200,
"sha256": "72d41a3fe5dcbf91eb33c1682cd4db5ccb220160c35e58e310ead96ac39ac600",
"entries": 2017
},
"warm": [
{
"concurrency": 1,
"requests": 1,
"p50Ms": 35981.36801400001,
"p95Ms": 35981.36801400001,
"p50FirstBodyMs": 39.2833660000033,
"requestsPerSecond": 0.027762249951141002,
"cpuMs": 35640,
"peakProcessTreeRssBytes": 74838016,
"bytes": 13045723
}
]
},
"memoryReductionPercent": 89.25829247357345,
"changes": [
"Default MongoDB wire compression changed from zstd to Snappy; explicit URI compressor settings respected",
"Projected MongoDB cursor with 128-entry batches; removed silent 100000-file truncation",
"One 32 KiB copy buffer reused per ZIP"
],
"validation": {
"raceTests": "ZIP streaming/cancellation, multi-batch archive contents and exclusions, second-provider lifecycle: passed (302.324s server package)",
"vet": "passed",
"live": {
"hostname": "beta.durieux.me",
"checks": 133,
"temporaryKeyRevoked": true,
"testRepositoryRemoved": true,
"image": "sha256:6817b036c77e372bbeb5fb2570e3df044af920d66fc51cbdb44536f866c92d28"
}
},
"limitations": [
"ZIP central directory and duplicate-name set still scale with entry count",
"Individual transformation workers still have their existing input and memory limits",
"Explicitly configuring zstd in the MongoDB URI can reintroduce its encoder memory costs"
]
}