Skip to content

performance.md

Latest commit

 

History

History
105 lines (76 loc) · 3.12 KB

performance.md

File metadata and controls

105 lines (76 loc) · 3.12 KB
Raw

Performance Baseline

This document provides a methodology for benchmarking Pebble CMS and reference performance characteristics.

Methodology

Tool: oha (or wrk / hey as alternatives)

Reference dataset:

  • 100 published posts with tags and markdown content
  • 20 published pages
  • 50 tags
  • 10 media files
  • SQLite database ~2 MB

Environment: Single-core VPS, 1 GB RAM, SSD storage (representative of a $5/mo cloud instance).

Benchmark Commands

# Homepage
oha -n 10000 -c 50 http://127.0.0.1:8080/

# Post page
oha -n 10000 -c 50 http://127.0.0.1:8080/posts/example-post

# Posts listing (paginated)
oha -n 10000 -c 50 http://127.0.0.1:8080/posts

# Search
oha -n 5000 -c 25 "http://127.0.0.1:8080/search?q=rust"

# API: list posts
oha -n 10000 -c 50 http://127.0.0.1:8080/api/v1/posts

# Health check
oha -n 10000 -c 50 http://127.0.0.1:8080/health

Reference Performance Table

Results will vary by hardware. These are reference numbers on a 1-core, 1 GB RAM VPS:

Endpoint RPS p50 (ms) p99 (ms) Notes
GET /health — — — Baseline
GET / — — — Homepage with hero + posts
GET /posts — — — Paginated listing
GET /posts/:slug — — — Single post render
GET /search?q=term — — — FTS5 search
GET /api/v1/posts — — — JSON API

Run the benchmarks on your target hardware to fill in these values.

Memory Profile

Metric Value
RSS at startup ~10-15 MB
RSS under load (50 concurrent) ~25-40 MB
Peak RSS ~50 MB

Pebble's memory footprint is intentionally small. SQLite operates within the configured cache_size (~64 MB page cache) and mmap_size (128 MB), but actual RSS depends on working set size.

Binary Size

Build Size
Debug ~60-80 MB
Release (with LTO + strip) ~8-12 MB

Release profile settings in Cargo.toml:

[profile.release]
lto = true
codegen-units = 1
strip = true
panic = "abort"

Cold Start Time

Pebble starts in under 100ms on typical hardware. The startup sequence:

  1. Load and validate config (~1ms)
  2. Open SQLite connection pool (~5ms)
  3. Run pending migrations (if any)
  4. Rebuild FTS index (~10-50ms depending on content volume)
  5. Bind TCP listener and serve

SQLite Tuning Rationale

Pebble applies the following PRAGMA settings for optimal performance:

Setting Value Rationale
journal_mode WAL Concurrent reads during writes
synchronous NORMAL Safe with WAL, faster than FULL
busy_timeout 5000ms Wait instead of failing on lock contention
journal_size_limit 64 MB Cap WAL file growth
mmap_size 128 MB Memory-mapped I/O for faster reads
cache_size -65536 (64 MB) In-memory page cache
foreign_keys ON Enforce referential integrity

These defaults work well for personal blogs with up to thousands of posts. For high-traffic deployments, consider placing the database on SSD storage and increasing cache_size.