{ "schemaVersion": 2, "updatedAt": "2026-08-11T14:21:08Z", "publicationPolicy": { "summary": "Only synthetic, reproducible benchmark evidence is published. Customer data and production traffic are never used.", "isolation": "The load generator, seed data and result storage run in a separate template-derived Railway project and are not part of the installable RustyAuth template.", "promotion": "A result becomes public only after its resource limits, dataset, workload, release commit, image digests and retained evidence pass catalogue validation." }, "programs": [ { "id": "single-realm-capacity", "title": "Single-realm capacity", "state": "active", "summary": "Measures how many registered accounts, valid sessions, active authenticated users and passkey sign-ins one supported single-writer realm can serve within explicit latency and reliability gates.", "schedule": [ "A fixed-rate regression run after every release", "A breakpoint ladder and one-hour soak once each month", "A manual run after authentication, gateway, datastore or deployment changes" ], "resourceTiers": [ { "name": "Starter", "api": "1 vCPU / 512 MB", "sableDb": "1 vCPU / 1 GB", "datasetAccounts": 10000 }, { "name": "Enterprise boundary", "api": "1 vCPU / 1 GB", "sableDb": "4 vCPU / 4 GB", "datasetAccounts": 10000 }, { "name": "Growth", "api": "2 vCPU / 1 GB", "sableDb": "2 vCPU / 2 GB", "datasetAccounts": 100000 }, { "name": "Scale", "api": "4 vCPU / 2 GB", "sableDb": "4 vCPU / 4 GB", "datasetAccounts": 1000000 } ], "userProfiles": [ { "name": "Light SaaS", "requestsPerMinute": 2, "description": "Occasional account and session-backed application activity." }, { "name": "Typical application", "requestsPerMinute": 6, "description": "A signed-in product session making a request about every ten seconds." }, { "name": "Heavy administration", "requestsPerMinute": 20, "description": "An operator or workflow making sustained authenticated calls." } ], "gates": [ "Unexpected request failures below 0.1% and no unplanned 5xx responses", "Mixed end-to-end latency below 300 ms p95 and 750 ms p99", "Application latency below 150 ms p95 / 400 ms p99 and accumulated SableDB below 150 ms p95 / 250 ms p99", "The measured phase completes its exact scheduled arrival count; excluded warm-up drops remain disclosed", "Complete passkey ceremonies remain below 750 ms p95 and 1,500 ms p99 under operating load", "A one-hour operating-rate soak completes without readiness loss, restart or out-of-memory event", "Published user capacity retains 30% measured throughput headroom" ], "decisionGuide": { "headline": "One measured Railway realm retained an exact 2,400 mixed authenticated RPS boundary pass; its one-hour operating-rate qualification remains incomplete.", "measured": "A 10,000-account single-writer realm passed an exact two-minute 2,400 RPS boundary. The exact 3,200 RPS run was the first tested failure. A companion passkey run completed 30 ceremonies under 1,680 RPS background load. The planned soak stopped after 37 minutes 40.7 seconds and did not run recovery.", "inferred": "Applying 30 percent planning headroom to the short-window boundary gives a candidate 1,680 RPS rate and candidate active-user projections; these are not published operating capacity.", "notDemonstrated": "A one-hour operating-rate soak and recovery, 50,000 or 100,000 stored accounts, multi-region active-active writes, shared Fleet bottlenecks and hyperscale operation remain unqualified.", "scaleStrategy": "Use the last fully published 800 RPS baseline and its 560 RPS headroom rate for supported planning until a fresh one-hour enterprise soak and recovery are completed. Treat the 2,400 RPS boundary and 1,680 RPS headroom calculation as informational." }, "enterpriseProfile": { "name": "High-traffic product journey v2", "state": "published", "mix": [ { "operation": "Session-backed account read", "percent": 60 }, { "operation": "User token mint", "percent": 20 }, { "operation": "Passkey inventory read", "percent": 15 }, { "operation": "Public signing-key discovery", "percent": 5 } ], "phases": [ "Warm-up and connection establishment", "Exact 2,400 RPS sustained qualification", "Exact 3,200 RPS first-failure confirmation", "One-hour 1,680 RPS operating-rate soak", "Complete passkey ceremonies under the soak", "Recovery and readiness verification" ], "timing": "Secret-gated Server-Timing separates the external path, API application work and every API-to-SableDB command or pipeline." } }, { "id": "fleet-analytics-query", "title": "Fleet Analytics query qualification", "state": "active", "summary": "Measures bounded organization-level Analytics queries at the pinned medium qualification tier.", "schedule": [ "Continuous assurance on the current main revision", "Repeated before a supported release is promoted" ], "resourceTiers": [], "userProfiles": [], "gates": [ "A 28-day organization query completes within 2 seconds p95 at the medium tier" ] } ], "reports": [ { "id": "single-realm-enterprise-2026-08-11", "programId": "single-realm-capacity", "title": "Enterprise mixed-journey Railway boundary and soak", "status": "informational", "qualification": "Retained two-minute boundary, first-failure and passkey evidence; one-hour soak stopped at 37m40.7s", "observedAt": "2026-08-11T14:12:49Z", "release": "1.0.0", "commit": "10b19eb48dc22e6271acea046c3da1ebaec7ad63", "environment": "Railway: API 1 vCPU / 1 GB; SableDB 4 vCPU / 4 GB; public Railway edge", "methodologyVersion": "single-realm-enterprise-v2", "summary": "A 10,000-account single-writer realm passed the exact two-minute mixed-workload boundary at 2,400 RPS. The exact 3,200 RPS test was the first failure. Thirty passkey ceremonies passed under 1,680 RPS background load, but the planned one-hour soak was stopped after 37 minutes 40.7 seconds, did not run recovery and recorded 548 ms Railway API-edge p95. The 1,680 RPS operating rate is therefore a candidate, not a qualified production claim.", "imageDigests": { "api": "sha256:e5aa606f94f2657bb41a6043f8360bf7290025c9e635ecd2159e043c1fa76c2f", "dashboard": "sha256:0409dccea72e6c7731c08a6324cabe393a785f81e28606cebcb7404832bf8962", "sableDb": "sha256:bbccbe52b7915a2d73b8cb0f9e12c706a5c733fd492432d341f66ef8d0b79051" }, "dataset": [ { "key": "mixed_requests_at_qualified_boundary", "label": "Measured mixed requests at 2,400 RPS", "value": 288140, "unit": "requests" }, { "key": "qualified_boundary_duration", "label": "Exact boundary window", "value": 120, "unit": "seconds" }, { "key": "passkey_ceremonies_under_load", "label": "Complete passkey sign-ins under load", "value": 30, "unit": "sign-ins" } ], "results": [ { "key": "registered_accounts", "label": "Registered accounts", "value": 10000, "unit": "accounts", "threshold": "10,000 seeded synthetic accounts", "outcome": "pass" }, { "key": "valid_sessions", "label": "Valid sessions", "value": 10000, "unit": "sessions", "threshold": "Equal to registered accounts", "outcome": "pass" }, { "key": "sustainable_authenticated_rps", "label": "Qualified mixed authenticated throughput", "value": 2400, "unit": "RPS", "threshold": "Every strict v2 gate passes for two measured minutes", "outcome": "pass" }, { "key": "candidate_operating_rps", "label": "Candidate rate with 30% headroom", "value": 1680, "unit": "RPS", "threshold": "Planning calculation pending one-hour soak and recovery", "outcome": "observe" }, { "key": "first_failing_rps", "label": "First tested failing rate", "value": 3200, "unit": "RPS", "threshold": "Scheduled-throughput and 150 ms internal p95 gates fail closed", "outcome": "observe" }, { "key": "candidate_typical_active_users", "label": "Candidate typical simultaneously active users", "value": 16800, "unit": "users", "threshold": "Unqualified projection at six requests/minute", "outcome": "observe" }, { "key": "signin_rps", "label": "Passkey sign-in companion rate", "value": 0.1, "unit": "sign-ins/s", "threshold": "6 complete ceremonies/minute under 1,680 RPS background", "outcome": "pass" }, { "key": "signin_p95_ms", "label": "Passkey sign-in p95 under load", "value": 221.1, "unit": "ms", "threshold": "< 750 ms", "outcome": "pass" }, { "key": "signin_p99_ms", "label": "Passkey sign-in p99 under load", "value": 310.75, "unit": "ms", "threshold": "< 1,500 ms", "outcome": "pass" }, { "key": "mixed_authenticated_p95_ms", "label": "2,400 RPS end-to-end p95", "value": 173.293139, "unit": "ms", "threshold": "< 300 ms", "outcome": "pass" }, { "key": "mixed_authenticated_p99_ms", "label": "2,400 RPS end-to-end p99", "value": 251.553855, "unit": "ms", "threshold": "< 750 ms", "outcome": "pass" }, { "key": "external_path_p95_ms", "label": "Public edge and network p95", "value": 55.297302, "unit": "ms", "threshold": "Observed separately from server work", "outcome": "observe" }, { "key": "application_p95_ms", "label": "Application p95", "value": 142.095, "unit": "ms", "threshold": "< 150 ms", "outcome": "pass" }, { "key": "sable_p95_ms", "label": "Accumulated SableDB p95", "value": 142.01925, "unit": "ms", "threshold": "< 150 ms", "outcome": "pass" }, { "key": "qualified_boundary_failure_percent", "label": "Qualified-boundary failure rate", "value": 0.032623, "unit": "%", "threshold": "< 0.1%", "outcome": "pass" }, { "key": "qualified_boundary_warmup_dropped_iterations", "label": "Excluded connection-warm-up drops", "value": 1386, "unit": "iterations", "threshold": "Outside the measured phase; measured count must still meet 288,000", "outcome": "observe" }, { "key": "failed_requests", "label": "Failed requests at qualified boundary", "value": 94, "unit": "requests", "threshold": "< 0.1% of scheduled requests", "outcome": "pass" }, { "key": "unplanned_5xx", "label": "Unplanned server 5xx", "value": 0, "unit": "responses", "threshold": "Exactly zero", "outcome": "pass" }, { "key": "attempted_soak_rps", "label": "Attempted partial-soak rate", "value": 1680, "unit": "RPS", "threshold": "One-hour soak required", "outcome": "observe" }, { "key": "partial_soak_duration_seconds", "label": "Observed partial-soak duration", "value": 2260.7, "unit": "seconds", "threshold": "3,600 seconds plus recovery required", "outcome": "fail" }, { "key": "partial_soak_completed_iterations", "label": "Completed iterations including warm-up", "value": 3802109, "unit": "iterations", "threshold": "Diagnostic only; no terminal summary", "outcome": "observe" }, { "key": "partial_soak_logged_http2_cancels", "label": "Logged HTTP/2 CANCEL warnings", "value": 284, "unit": "log lines", "threshold": "Diagnostic only; terminal failure rate unavailable", "outcome": "observe" }, { "key": "partial_soak_api_edge_p95_ms", "label": "Railway API-edge p95 in partial window", "value": 548, "unit": "ms", "threshold": "< 300 ms", "outcome": "fail" }, { "key": "partial_soak_api_5xx", "label": "Railway API 5xx in partial window", "value": 0, "unit": "responses", "threshold": "Exactly zero", "outcome": "pass" } ], "capacityModels": [ { "profile": "Light SaaS", "requestsPerMinute": 2, "activeUsers": 50400, "basis": "1,680 candidate RPS × 60 ÷ 2" }, { "profile": "Typical application", "requestsPerMinute": 6, "activeUsers": 16800, "basis": "1,680 candidate RPS × 60 ÷ 6" }, { "profile": "Heavy administration", "requestsPerMinute": 20, "activeUsers": 5040, "basis": "1,680 candidate RPS × 60 ÷ 20" } ], "realmScaling": { "measuredRealmCells": 1, "model": "linear-independent-cells", "formula": "equivalent independent realm cells × candidate 1,680 RPS", "limitations": [ "Only one realm cell was measured; rows above one are planning extrapolations.", "Equivalent cells require the same images, resources, dataset class and workload mix.", "Shared Fleet, ingress, regional, backup and observability limits are not included.", "The cell model does not create multi-writer consistency inside one realm.", "The candidate rate did not complete the one-hour soak or recovery gate and is not supported production capacity." ] }, "charts": [ { "id": "mixed-boundary-latency", "title": "The mixed journey clears 2,400 RPS and bends at 3,200 RPS", "description": "Exact two-minute public Railway qualification windows. The 2,400 RPS point passes every gate; 3,200 RPS is retained as the first scheduled-throughput and internal-latency failure.", "xUnit": "RPS", "yUnit": "ms", "series": [ { "name": "end-to-end", "points": [ { "x": 2400, "y": 173.293139 }, { "x": 3200, "y": 295.903691 } ] }, { "name": "application", "points": [ { "x": 2400, "y": 142.095 }, { "x": 3200, "y": 236.73965 } ] }, { "name": "SableDB", "points": [ { "x": 2400, "y": 142.01925 }, { "x": 3200, "y": 236.64055 } ] }, { "name": "external path", "points": [ { "x": 2400, "y": 55.297302 }, { "x": 3200, "y": 81.76177 } ] } ] } ], "confidence": { "measured": "Pinned Railway images and resources; 10,000 accounts and sessions; exact 2,400 RPS two-minute pass; exact 3,200 RPS first failure; 30 passkey ceremonies under 1,680 RPS background load; 37m40.7s incomplete soak; real public edge and private SableDB path.", "inferred": "Active-user and multi-realm figures apply the disclosed activity model and linear equivalent-cell assumptions to the unqualified 1,680 RPS candidate rate.", "notProven": "A one-hour operating-rate soak and recovery, 50,000 or 100,000 stored accounts, high-rate login from distributed source IPs, HA failover, multi-region writes, shared Fleet limits and hyperscale operation remain separate gates." }, "evidence": [ { "label": "Informational enterprise report", "url": "https://github.com/rusty-auth/rustyauth/blob/main/release-evidence/benchmarks/enterprise-20260811T132328Z/REPORT.md" }, { "label": "Railway wind-down record", "url": "https://github.com/rusty-auth/rustyauth/blob/main/release-evidence/benchmarks/enterprise-20260811T132328Z/WIND_DOWN.md" }, { "label": "Reviewed metrics and deployment artifacts", "url": "https://github.com/rusty-auth/rustyauth/blob/main/release-evidence/benchmarks/enterprise-20260811T132328Z/reviewed-evidence.json" }, { "label": "Raw final-run artifacts", "url": "https://github.com/rusty-auth/rustyauth/tree/main/release-evidence/benchmarks/enterprise-20260811T132328Z/raw-runs" }, { "label": "Captured Railway metrics", "url": "https://github.com/rusty-auth/rustyauth/tree/main/release-evidence/benchmarks/enterprise-20260811T132328Z/railway-metrics" } ] }, { "id": "single-realm-starter-2026-08-10", "programId": "single-realm-capacity", "title": "Starter single-realm Railway baseline", "status": "passed", "qualification": "Reviewed Railway production baseline", "observedAt": "2026-08-10T15:06:24Z", "release": "1.0.0", "commit": "6a73d1ce366032ac13e65da45e5c8abecabcb1f4", "environment": "Railway Starter: API 1 vCPU / 512 MB; SableDB 1 vCPU / 1 GB", "methodologyVersion": "single-realm-v1", "summary": "A template-derived single-writer realm passed every configured step through 800 authenticated requests per second, so this run establishes a lower bound rather than an absolute breakpoint. Across 141,764 authenticated reads and 31 complete passkey sign-ins there were zero failed requests, zero unplanned 5xx responses and zero dropped iterations.", "imageDigests": { "api": "sha256:507a5e5da1660ea780f2b894b2daaae99b8af6b1373ab0b6401bf5b6bf6af112", "dashboard": "sha256:abc9f12bab4c12f62752697fbda39b2dc5d0c461fd335420a87c8809e855d005", "sableDb": "sha256:83b665171d5994ac7f0d2456bb53a9ab2cc56b69763008d37994de84e2752a3a" }, "dataset": [ { "key": "authenticated_read_requests", "label": "Authenticated reads", "value": 141764, "unit": "requests" }, { "key": "passkey_ceremonies", "label": "Complete passkey sign-ins", "value": 31, "unit": "sign-ins" } ], "results": [ { "key": "registered_accounts", "label": "Registered accounts", "value": 10000, "unit": "accounts", "threshold": "10,000 seeded synthetic accounts", "outcome": "pass" }, { "key": "valid_sessions", "label": "Valid sessions", "value": 10000, "unit": "sessions", "threshold": "Equal to registered accounts", "outcome": "pass" }, { "key": "sustainable_authenticated_rps", "label": "Authenticated read capacity", "value": 800, "unit": "RPS", "threshold": "All gates at 800 RPS; breakpoint not reached", "outcome": "pass" }, { "key": "supported_typical_active_users", "label": "Typical active users", "value": 5600, "unit": "users", "threshold": "30% headroom at 6 requests/minute", "outcome": "pass" }, { "key": "signin_rps", "label": "Passkey sign-in rate", "value": 0.1033047791, "unit": "sign-ins/s", "threshold": "Below the 10 probes/minute source limit", "outcome": "pass" }, { "key": "signin_p95_ms", "label": "Passkey sign-in p95", "value": 118, "unit": "ms", "threshold": "< 750 ms", "outcome": "pass" }, { "key": "signin_p99_ms", "label": "Passkey sign-in p99", "value": 129.1, "unit": "ms", "threshold": "< 1,500 ms", "outcome": "pass" }, { "key": "authenticated_read_p95_ms", "label": "800 RPS read p95", "value": 76.456158, "unit": "ms", "threshold": "< 250 ms", "outcome": "pass" }, { "key": "authenticated_read_p99_ms", "label": "800 RPS read p99", "value": 130.792856, "unit": "ms", "threshold": "< 750 ms", "outcome": "pass" }, { "key": "failed_requests", "label": "Failed requests", "value": 0, "unit": "requests", "threshold": "< 0.1%", "outcome": "pass" }, { "key": "api_peak_vcpu", "label": "API peak CPU", "value": 0.1401467, "unit": "vCPU", "threshold": "≤ 1 vCPU", "outcome": "pass" }, { "key": "api_peak_memory_mb", "label": "API peak memory", "value": 36.557553664, "unit": "MB", "threshold": "≤ 512 MB", "outcome": "pass" }, { "key": "sable_peak_vcpu", "label": "SableDB peak CPU", "value": 0.3840858333, "unit": "vCPU", "threshold": "≤ 1 vCPU", "outcome": "pass" }, { "key": "sable_peak_memory_mb", "label": "SableDB peak memory", "value": 133.722800128, "unit": "MB", "threshold": "≤ 1,024 MB", "outcome": "pass" } ], "capacityModels": [ { "profile": "Light SaaS", "requestsPerMinute": 2, "activeUsers": 16800, "basis": "800 RPS × 60 × 70% ÷ 2" }, { "profile": "Typical application", "requestsPerMinute": 6, "activeUsers": 5600, "basis": "800 RPS × 60 × 70% ÷ 6" }, { "profile": "Heavy administration", "requestsPerMinute": 20, "activeUsers": 1680, "basis": "800 RPS × 60 × 70% ÷ 20" } ], "charts": [ { "id": "authenticated-read-latency", "title": "Latency remains flat as traffic rises", "description": "End-to-end public Railway latency at each fixed authenticated-read arrival rate. The 800 RPS point is a measured lower bound, not the breakpoint.", "xUnit": "RPS", "yUnit": "ms", "series": [ { "name": "p50", "points": [ { "x": 25, "y": 25.8239975 }, { "x": 50, "y": 25.028347 }, { "x": 100, "y": 24.96317 }, { "x": 200, "y": 24.531987 }, { "x": 400, "y": 24.624215 }, { "x": 800, "y": 25.305971 } ] }, { "name": "p95", "points": [ { "x": 25, "y": 74.4825078 }, { "x": 50, "y": 65.139034 }, { "x": 100, "y": 66.949184 }, { "x": 200, "y": 73.31633 }, { "x": 400, "y": 75.495421 }, { "x": 800, "y": 76.456158 } ] }, { "name": "p99", "points": [ { "x": 25, "y": 108.21646818 }, { "x": 50, "y": 99.226223 }, { "x": 100, "y": 104.14369 }, { "x": 200, "y": 113.83114789 }, { "x": 400, "y": 112.36823 }, { "x": 800, "y": 130.792856 } ] } ] } ], "confidence": { "measured": "Pinned Railway resources and images; real public edge; production session validation; zero failures through 800 RPS.", "inferred": "Active-user figures apply the disclosed request-frequency model and reserve 30% throughput headroom.", "notProven": "The absolute single-realm breakpoint, one-hour mixed-workload soak, multi-region cells and Facebook-scale Fleet cardinality remain separate evidence gates." }, "evidence": [ { "label": "Reviewed report", "url": "https://github.com/rusty-auth/rustyauth/blob/main/release-evidence/benchmarks/starter-20260810T145154Z/REPORT.md" }, { "label": "Reviewed metrics and deployment artifacts", "url": "https://github.com/rusty-auth/rustyauth/blob/main/release-evidence/benchmarks/starter-20260810T145154Z/reviewed-evidence.json" }, { "label": "Raw k6 summaries", "url": "https://github.com/rusty-auth/rustyauth/blob/main/release-evidence/benchmarks/starter-20260810T145154Z/raw-k6-summaries.json" }, { "label": "Runner source checksums", "url": "https://github.com/rusty-auth/rustyauth/blob/main/release-evidence/benchmarks/starter-20260810T145154Z/source-SHA256SUMS" } ] }, { "id": "fleet-analytics-medium-2026-08-09", "programId": "fleet-analytics-query", "title": "Medium-tier 28-day organization query", "status": "passed", "qualification": "Pinned local source qualification", "observedAt": "2026-08-09T00:00:00Z", "release": "1.0.0 source", "commit": "source-qualified-2026-08-09", "environment": "Pinned local SableDB and GreptimeDB topology", "methodologyVersion": "fleet-analytics-v1", "summary": "Twenty organization queries were measured after seeding 1,000 realms and 8,064,000 five-minute metric rows.", "dataset": [ { "key": "realms", "label": "Realms", "value": 1000, "unit": "realms" }, { "key": "metric_rows", "label": "Metric rows", "value": 8064000, "unit": "rows" }, { "key": "query_samples", "label": "Query samples", "value": 20, "unit": "queries" } ], "results": [ { "key": "organization_query_p95_ms", "label": "Organization query p95", "value": 239.5325, "unit": "ms", "threshold": "≤ 2,000 ms", "outcome": "pass" } ], "evidence": [ { "label": "Release-readiness evidence", "url": "https://github.com/rusty-auth/rustyauth/blob/main/docs/RELEASE_READINESS.md" }, { "label": "Fleet Analytics runbook", "url": "https://github.com/rusty-auth/rustyauth/blob/main/docs/FLEET_ANALYTICS_RUNBOOK.md" } ] } ] }