Benchmarks & Profiling - Infra-Index Platform

Performance Benchmarks & Latency Profiling

Empirical latency, throughput, and cache efficiency metrics measured across global regions.

Real Production Latency Benchmark (Live Measurements)

Benchmark tests executed against Vercel Edge Serverless backend connected to Upstash Redis and Neon PostgreSQL clusters:

Upstash Redis L1 Hit

0.35s ~ 0.41s

End-to-end HTTP response time through Vercel Edge and global Upstash Redis cluster.

Neon DB Cold Query

0.82s ~ 0.95s

Database fallback time across 32,000+ time-series records with Composite B-Tree indexes.

Previous Unindexed Cold Scan

15.46s

95% performance improvement achieved via (model, provider, region, timestamp) composite indexing.

Endpoint Response Time Comparison Table

Endpoint & Menu Category	Payload Size	Item Count	Cached Latency (L1)	DB Cold Query	Cache Hit Rate
`/api/v1/data/list?menu_id=gpu`	29.0 KB	159 Models	0.412s	0.821s	99.9%
`/api/v1/data/list?menu_id=cpu`	2.9 KB	23 Models	0.354s	0.958s	99.9%
`/api/v1/data/list?menu_id=storage`	3.1 KB	90 Providers	0.388s	0.672s	99.8%
`/api/v1/data/list?menu_id=baremetal`	2.7 KB	220 Servers	0.369s	0.708s	99.8%
`/api/v1/health/ping`	0.1 KB	1 Health Obj	0.045s	0.082s	100.0%

Database Index Optimization Profile

-- Neon PostgreSQL Indexing Benchmark
-- Table: tbl_cpu_prc_hist (31,906 rows)

-- BEFORE INDEX (Full Sequential Scan):
-- Execution Time: 15,463 ms | Planning Time: 12.4 ms | Memory Sort: 14.8 MB

-- AFTER COMPOSITE INDEX CREATION:
CREATE INDEX idx_cpu_mdl_prv_region_ts ON "tbl_cpu_prc_hist" (cpu_mdl, prv_id, region, ts DESC);
-- Execution Time: 71.9 ms (99.5% reduction in PostgreSQL execution overhead)