Verified Benchmarks & CI Size Guardrails

DioxusPress follows a strict "measure first, optimize by verified impact, and guard in CI" discipline. Every metric below comes directly from built-in scripts (./scripts/bench_ssr.sh and ./scripts/check_wasm_size.sh) that you can reproduce locally in minutes.

1. Binary, Container Image & Memory Footprint

Measured across our multi-site production stack (debian:trixie-slim runtime images):

Component / Tier	Binary / Image Size	Cold Start to `/healthz 200`	Idle Memory (RSS)	Typical Workload
`dioxuspress-gateway` (Pingora TLS SNI proxy)	`143 MB` image	`< 50 ms`	`6.4 MiB`	Multi-domain TLS termination, HTTP/2, and `301` host redirects
`basic` Zero-DB Tier (`--no-default-features --features server`)	`34.2 MiB` bin / `269 MB` image	`547 ms`	`91.7 MiB` (includes Jieba dictionary + Tantivy mmap)	Open-source docs & blogs (e.g. `dioxuspress.dev`)
`accounts` / `full` Tier	`47.4 MiB` bin / `319 MB` image	`549 ms`	`95.6 MiB` (~`459 MiB` live RSS serving 2,200+ indexed docs & courses)	Fullstack community & paid course portals (e.g. `rustineverything.app`)

2. First-Paint SSR, Conditional `ETag` (`304`), and Tantivy Search Latency

Measured on Apple Silicon loopback (RUST_LOG=warn, --release, ./scripts/bench_ssr.sh, N=200 serial samples and c=16 concurrent load with N=1,600 requests):

Endpoint (`basic` Tier)	Status	P50 Latency	P99 Latency	`c=16` Throughput	Notes
Home Page SSR (`GET /`)	`200 OK`	`2.79 ms`	`3.57 ms`	`1,832 QPS`	Full layout, navbar, feature grid, and SEO metadata
Docs Page SSR (`GET /docs/guide/quickstart`)	`200 OK`	`4.04 ms`	`5.05 ms`	`1,236 QPS`	Live Markdown parse + 3-level sidebar + breadcrumbs + VirtualDOM SSR
English Companion SSR (`GET /docs/guide/quickstart?lang=en`)	`200 OK`	`4.15 ms`	`6.67 ms`	`1,219 QPS`	`index.en.md` companion resolution + `hreflang` alternates
Anonymous Conditional Revalidation (`If-None-Match`)	`304 Not Modified`	`0.096 ms`	`0.151 ms`	`13,636 QPS`	Skips Markdown & VirtualDOM SSR completely (41x faster P50, 10.3x QPS)
Atom Feed (`GET /feed.xml`)	`200 OK`	`0.182 ms`	`0.310 ms`	`11,785 QPS`	Served from `DirListingCache`
Tantivy Full-Text Search (`POST /api/search/query`)	`200 OK`	`0.314 ms`	`0.425 ms`	`9,215 QPS`	BM25 ranking + bilingual Jieba tokenization + highlighted snippets

💡TIP

Strict Privacy Isolation Against Cache Cross-Contamination

conditional_cache_mw enforces a fail-closed cookie allowlist: weak ETag caching and 304 short-circuiting apply only to anonymous GET/HEAD requests carrying no Authorization header and at most display-only cookies (site_lang, site_theme). Any request carrying session or oauth_pkce cookies — or targeting /admin, /me, /forum, /topics, /course, or /api — receives Cache-Control: private, no-store with ETag stripped.

3. Zero-Downtime `CONTENT_DIR` Hot Reload (`SIGHUP`)

When external content is mounted at CONTENT_DIR, running docker kill --signal=HUP dioxuspress-app refreshes content in-place without restarting the container or dropping active TCP connections. Verified under 8-worker continuous load (1,782 concurrent requests):

Error rate during reload: 0 non-200 responses;
SSR prose propagation: 3.96 ms;
Background incremental Tantivy reindex: 272 ms (offloaded via spawn_blocking behind an async writer mutex while search readers remain 100% lock-free).

4. Frontend WebAssembly Bundle Size & CI Guardrail

Root Cargo.toml defines [profile.wasm-release] (opt-level = "z", lto = "fat", codegen-units = 1, strip = "debuginfo"), and GitHub Actions CI enforces upper bounds on every pull request via ./scripts/check_wasm_size.sh:

Client Build Tier	Raw `.wasm` (pre-`wasm-opt`)	`gzip -9` Transfer Size	`brotli -q 11` Transfer Size	CI Gzip Ceiling
`basic` (Docs & Blog Hydration)	`5.52 MiB` (`5,790,105 B`)	`1.54 MiB`	`1.04 MiB`	`1.86 MiB` (`1,950,000 B`)
`full` (Community, Courses & Admin)	`7.59 MiB` (`7,958,284 B`)	`2.07 MiB`	`1.35 MiB` (~`1.19 MiB` after `dx bundle` `wasm-opt`)	`2.43 MiB` (`2,550,000 B`)

5. Reproduce Locally

# 1. Run full SSR, 304 conditional cache, RSS, and Tantivy benchmark across basic & accounts tiers
./scripts/bench_ssr.sh

# 2. Run the CI WebAssembly bundle size guardrail
./scripts/check_wasm_size.sh