HeliosDB-Full vs PostgreSQL 14 — Performance Comparison

HeliosDB-Full vs PostgreSQL 14 — Performance Comparison

Date: 2026-02-14 22:47:46
HeliosDB: v7.2.0 (compiled binary, port 5435)
PostgreSQL: 14.20 (Docker container, port 25432)
Table: 1000 rows, 5 columns (INT, TEXT, TEXT, INT, TEXT)
Protocol: PostgreSQL wire protocol (psycopg2)

Results

Test	Queries	HeliosDB avg	PostgreSQL avg	Winner
Point Lookups	2000	75us	132us	HeliosDB 1.8x FASTER
Full Scan SELECT	200	323us	859us	HeliosDB 2.7x FASTER
Aggregates	198	55us	148us	HeliosDB 2.7x FASTER
Single-Row INSERT	500	75us	481us	HeliosDB 6.4x FASTER
UPDATE WHERE	500	83us	466us	HeliosDB 5.6x FASTER
DELETE WHERE	200	70us	450us	HeliosDB 6.5x FASTER
Mixed OLTP	1000	82us	271us	HeliosDB 3.3x FASTER

HeliosDB Optimizations Tested

1. Point Lookup Fast Path (Imp 8): pk_index → O(1) lsm_get instead of full table scan
2. Group-Commit WAL (Imp 1): Batched write-ahead log for 26x internal write speedup
3. SQL Parse Cache (Imp 5): Cached AST for repeated query patterns
4. SELECT Overhead Reduction (Imp 2): into_iter(), skip UTF-8 alloc, Vec::with_capacity
5. Finer PerfTracer Spans (Imp 7): Diagnostic visibility into row processing

Notes

• Both databases accessed via psycopg2 over TCP loopback (127.0.0.1)
• HeliosDB uses LSM-tree storage; PostgreSQL uses heap + B-tree indexes
• PostgreSQL has years of production optimization; HeliosDB is research-grade
• Network round-trip and psycopg2 overhead are included in both measurements
• Both tables have PRIMARY KEY on id (equal conditions)
• HeliosDB point lookups use in-memory pk_index (populated during INSERT)
• PostgreSQL uses B-tree index on PRIMARY KEY