Visual Summary: SQLite rc3 vs MariaDB
Green favors SQLite rc3. Red favors MariaDB. Gray is within 3%.
Throughput and writes are better when higher. Avg flow and p95 are better when lower, so a green latency chip means SQLite was faster.
| Workload / concurrent requests | 4 | 16 | 32 | 64 |
|---|---|---|---|---|
| read-heavy 90% catalog browsing/search and 10% logged-in order-admin reads; no checkout writes. | throughput: 12.8% loweravg flow: 14.7% slowerp95: 18.6% slowerwrites: not in mix | throughput: 20.3% loweravg flow: 25.2% slowerp95: 5.5% fasterwrites: not in mix | throughput: 18.4% loweravg flow: 22.5% slowerp95: 20.6% fasterwrites: not in mix | throughput: 27.1% loweravg flow: 36.6% slowerp95: 20.7% fasterwrites: not in mix |
| balanced 50% catalog browsing/search, 20% add-to-cart flows, 20% checkout/order writes, 10% order-admin reads. | throughput: 6.2% higheravg flow: 5.9% fasterp95: 1.7% fasterwrites: 2.3% lower | throughput: 23.7% loweravg flow: 30.7% slowerp95: 1.0% slowerwrites: 30.0% lower | throughput: 24.5% loweravg flow: 31.6% slowerp95: 13.5% fasterwrites: 27.8% lower | throughput: 32.6% loweravg flow: 49.1% slowerp95: 13.6% fasterwrites: 32.3% lower |
| write-heavy 20% catalog browsing, 20% add-to-cart flows, 50% checkout/order writes, 10% order-admin reads. | throughput: 13.0% higheravg flow: 12.0% fasterp95: 7.6% fasterwrites: 9.0% higher | throughput: 15.5% loweravg flow: 17.8% slowerp95: 10.4% slowerwrites: 16.0% lower | throughput: 24.7% loweravg flow: 32.0% slowerp95: 33.9% slowerwrites: 26.9% lower | throughput: 32.3% loweravg flow: 48.7% slowerp95: 40.0% slowerwrites: 32.5% lower |
How The Test Was Conducted
The measured window starts after restore, server startup, login setup, and warm-up.
- Build one fresh WordPress site per selected database variant from the same downloaded WordPress package.
- Install the SQLite driver or configure MariaDB, install WooCommerce, and generate the same seeded 5GB WooCommerce fixture for each variant.
- Create a clean database snapshot for each variant after fixture generation.
- For every measured cell, restore that variant's clean snapshot before starting the measured case.
- Start local nginx and PHP-FPM for that one case, then prepare logged-in admin/customer sessions outside the measured timing window.
- Run a 10-second warm-up so first-request and cache effects do not dominate the measured result.
- Run the measured workload for 30 seconds at the configured number of concurrent workers: 4, 16, 32, or 64.
- The load generator did not literally open Chrome. It issued browser-equivalent HTTP requests for the same user journeys.
- Public browse flow: visit the shop/home page, open product pages, and perform catalog searches.
- Cart flow: open a product page, submit add-to-cart, and load the cart page.
- Checkout flow: add a product, open checkout, submit the WooCommerce order form with benchmark billing details, and verify that an order was created.
- Admin read flow: log in as an administrator and open WooCommerce order-admin screens.
- Record per-case request logs, full-flow timings, average HTTP request timings, write verification, failures, SQLite WAL size, SQLite lock messages, and MariaDB status counters.
- Repeat each workload/concurrent-request/variant cell three times and report medians.
What Was Compared
These definitions apply to every table below.
Database Variants
- MariaDB baseline: WordPress Core using the local MariaDB/InnoDB database.
- SQLite rc3: WordPress Core using SQLite Database Integration v3.0.0-rc.3.
- SQLite stable: WordPress Core using SQLite Database Integration v2.2.23 as the stable-line control.
- SQLite PR #378: WordPress Core using SQLite Database Integration PR #378, parser + lexer performance branch.
Workload Mixes
- read-heavy: 90% catalog browsing/search and 10% logged-in order-admin reads; no checkout writes.
- balanced: 50% catalog browsing/search, 20% add-to-cart flows, 20% checkout/order writes, 10% order-admin reads.
- write-heavy: 20% catalog browsing, 20% add-to-cart flows, 50% checkout/order writes, 10% order-admin reads.
How To Read The Numbers
Comparison sections state their baseline in the section heading. For throughput columns, positive is good for the left-hand SQLite variant and negative is bad. For latency columns, negative is good because lower latency is faster.
- Flows/s: completed user-flow loops per second. Higher is better.
- Avg flow: average full user-flow loop completion time. Lower is better.
- Avg HTTP request: average individual HTTP request completion time inside those flows. Lower is better.
- p50 / p75 / p95 / p99: full-flow duration percentiles. p50 is the median, p95 is the slowest 5% boundary, and p99 is the slowest 1% boundary. Lower is better.
- Write/s: verified write flows per second. Higher is better.
- Write p95: 95th percentile duration for write flows only. Lower is better.
- SQLite lock errors: logged SQLITE_BUSY, SQLITE_LOCKED, or database-is-locked messages. Zero does not mean zero waiting; it means no visible lock failures were logged.
Methodology Fixes In This Run
These methodology controls make the local request-load results comparable across variants and avoid counting setup work as measured latency.
- database snapshot restored before every measured case
- login/setup excluded from measured elapsed time
- concurrent warm-up before measured window
- three repetitions by default
- variants executed sequentially because the local VM did not have enough free disk for all active 5GB sites at once
- verified WooCommerce checkout/order write flows
- per-case request logs
- SQLite WAL sampled without pre-recording checkpoint
- database snapshots compressed to reduce benchmark artifact disk usage
- after each variant finished, its active 5GB site/database was deleted to preserve local disk space
SQLite rc3 vs MariaDB
Each row compares SQLite rc3 against MariaDB for the same workload and number of concurrent requests.
| Workload | Concurrent requests | Plain English | rc3 flows/s | MariaDB flows/s | SQLite throughput vs MariaDB | rc3 avg flow | MariaDB avg flow | SQLite avg flow time vs MariaDB | rc3 p95 | MariaDB p95 | SQLite p95 latency vs MariaDB | rc3 write/s | MariaDB write/s | SQLite write throughput vs MariaDB | rc3 write p95 | MariaDB write p95 | SQLite write p95 vs MariaDB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| read-heavy | 4 | throughput: SQLite was 12.8% lower; p95 latency: SQLite was 18.6% slower | 40.79 | 46.76 | -12.8% | 98 ms | 85 ms | +14.7% | 174 ms | 147 ms | +18.6% | 0.00 | 0.00 | ||||
| read-heavy | 16 | throughput: SQLite was 20.3% lower; p95 latency: SQLite was 5.5% faster | 37.51 | 47.08 | -20.3% | 424 ms | 339 ms | +25.2% | 810 ms | 857 ms | -5.5% | 0.00 | 0.00 | ||||
| read-heavy | 32 | throughput: SQLite was 18.4% lower; p95 latency: SQLite was 20.6% faster | 36.88 | 45.20 | -18.4% | 862 ms | 703 ms | +22.5% | 1,531 ms | 1,927 ms | -20.6% | 0.00 | 0.00 | ||||
| read-heavy | 64 | throughput: SQLite was 27.1% lower; p95 latency: SQLite was 20.7% faster | 32.24 | 44.25 | -27.1% | 1,947 ms | 1,425 ms | +36.6% | 3,264 ms | 4,115 ms | -20.7% | 0.00 | 0.00 | ||||
| balanced | 4 | throughput: SQLite was 6.2% higher; p95 latency: SQLite was 1.7% faster; write throughput: SQLite was 2.3% lower; write p95: SQLite was 7.3% faster | 17.99 | 16.94 | +6.2% | 221 ms | 235 ms | -5.9% | 661 ms | 673 ms | -1.7% | 3.97 | 4.06 | -2.3% | 720 ms | 777 ms | -7.3% |
| balanced | 16 | throughput: SQLite was 23.7% lower; p95 latency: SQLite was 1.0% slower; write throughput: SQLite was 30.0% lower; write p95: SQLite was 0.7% slower | 17.68 | 23.16 | -23.7% | 887 ms | 679 ms | +30.7% | 2,752 ms | 2,726 ms | +1.0% | 3.23 | 4.62 | -30.0% | 3,121 ms | 3,100 ms | +0.7% |
| balanced | 32 | throughput: SQLite was 24.5% lower; p95 latency: SQLite was 13.5% faster; write throughput: SQLite was 27.8% lower; write p95: SQLite was 11.8% faster | 18.86 | 25.00 | -24.5% | 1,642 ms | 1,248 ms | +31.6% | 4,371 ms | 5,053 ms | -13.5% | 3.72 | 5.15 | -27.8% | 4,781 ms | 5,420 ms | -11.8% |
| balanced | 64 | throughput: SQLite was 32.6% lower; p95 latency: SQLite was 13.6% faster; write throughput: SQLite was 32.3% lower; write p95: SQLite was 11.8% faster | 16.50 | 24.48 | -32.6% | 3,688 ms | 2,473 ms | +49.1% | 8,525 ms | 9,864 ms | -13.6% | 3.40 | 5.02 | -32.3% | 9,315 ms | 10,563 ms | -11.8% |
| write-heavy | 4 | throughput: SQLite was 13.0% higher; p95 latency: SQLite was 7.6% faster; write throughput: SQLite was 9.0% higher; write p95: SQLite was 11.4% faster | 11.34 | 10.03 | +13.0% | 349 ms | 396 ms | -12.0% | 703 ms | 761 ms | -7.6% | 5.55 | 5.10 | +9.0% | 725 ms | 818 ms | -11.4% |
| write-heavy | 16 | throughput: SQLite was 15.5% lower; p95 latency: SQLite was 10.4% slower; write throughput: SQLite was 16.0% lower; write p95: SQLite was 1.5% slower | 11.04 | 13.08 | -15.5% | 1,410 ms | 1,197 ms | +17.8% | 2,747 ms | 2,489 ms | +10.4% | 5.52 | 6.57 | -16.0% | 2,912 ms | 2,870 ms | +1.5% |
| write-heavy | 32 | throughput: SQLite was 24.7% lower; p95 latency: SQLite was 33.9% slower; write throughput: SQLite was 26.9% lower; write p95: SQLite was 36.4% slower | 11.51 | 15.29 | -24.7% | 2,702 ms | 2,046 ms | +32.0% | 5,758 ms | 4,299 ms | +33.9% | 5.55 | 7.60 | -26.9% | 6,247 ms | 4,580 ms | +36.4% |
| write-heavy | 64 | throughput: SQLite was 32.3% lower; p95 latency: SQLite was 40.0% slower; write throughput: SQLite was 32.5% lower; write p95: SQLite was 41.5% slower | 10.49 | 15.50 | -32.3% | 5,841 ms | 3,928 ms | +48.7% | 12,037 ms | 8,599 ms | +40.0% | 5.12 | 7.58 | -32.5% | 12,767 ms | 9,023 ms | +41.5% |
good for SQLite
bad for SQLite
within 3%
SQLite rc3 vs Stable SQLite 2.2.23
This isolates the release-candidate driver from the stable plugin line under the same workload and concurrency levels.
| Workload | Concurrent requests | Plain English | rc3 throughput vs stable | rc3 avg flow time vs stable | rc3 p95 latency vs stable | rc3 write throughput vs stable | rc3 write p95 vs stable | rc3 max WAL | Stable max WAL |
|---|---|---|---|---|---|---|---|---|---|
| read-heavy | 4 | throughput: SQLite was 0.9% higher; p95 latency: SQLite was 2.9% faster | +0.9% | -0.9% | -2.9% | 4.0 MB | 4.0 MB | ||
| read-heavy | 16 | throughput: about the same; p95 latency: about the same | -0.3% | +0.2% | +0.4% | 4.3 MB | 4.2 MB | ||
| read-heavy | 32 | throughput: about the same; p95 latency: SQLite was 1.0% slower | +0.1% | +0.2% | +1.0% | 4.9 MB | 6.1 MB | ||
| read-heavy | 64 | throughput: SQLite was 8.7% lower; p95 latency: SQLite was 6.1% slower | -8.7% | +9.0% | +6.1% | 12.7 MB | 6.6 MB | ||
| balanced | 4 | throughput: SQLite was 1.7% lower; p95 latency: SQLite was 4.0% slower; write throughput: SQLite was 2.0% higher; write p95: SQLite was 4.3% slower | -1.7% | +2.8% | +4.0% | +2.0% | +4.3% | 6.5 MB | 6.5 MB |
| balanced | 16 | throughput: SQLite was 7.6% lower; p95 latency: SQLite was 20.3% slower; write throughput: SQLite was 4.8% lower; write p95: SQLite was 9.0% slower | -7.6% | +8.8% | +20.3% | -4.8% | +9.0% | 80.7 MB | 43.7 MB |
| balanced | 32 | throughput: SQLite was 47.0% higher; p95 latency: SQLite was 25.7% faster; write throughput: SQLite was 71.9% higher; write p95: SQLite was 40.9% faster | +47.0% | -30.8% | -25.7% | +71.9% | -40.9% | 120.9 MB | 82.7 MB |
| balanced | 64 | throughput: SQLite was 44.2% higher; p95 latency: SQLite was 29.9% faster; write throughput: SQLite was 78.8% higher; write p95: SQLite was 35.9% faster | +44.2% | -27.4% | -29.9% | +78.8% | -35.9% | 90.3 MB | 68.3 MB |
| write-heavy | 4 | throughput: SQLite was 5.0% lower; p95 latency: SQLite was 7.0% slower; write throughput: SQLite was 1.6% lower; write p95: SQLite was 5.4% slower | -5.0% | +5.7% | +7.0% | -1.6% | +5.4% | 9.6 MB | 10.4 MB |
| write-heavy | 16 | throughput: SQLite was 65.3% higher; p95 latency: SQLite was 45.1% faster; write throughput: SQLite was 85.9% higher; write p95: SQLite was 46.3% faster | +65.3% | -37.8% | -45.1% | +85.9% | -46.3% | 116.0 MB | 98.1 MB |
| write-heavy | 32 | throughput: SQLite was 68.4% higher; p95 latency: SQLite was 37.4% faster; write throughput: SQLite was 78.7% higher; write p95: SQLite was 36.4% faster | +68.4% | -36.2% | -37.4% | +78.7% | -36.4% | 208.0 MB | 153.1 MB |
| write-heavy | 64 | throughput: SQLite was 65.6% higher; p95 latency: SQLite was 22.7% faster; write throughput: SQLite was 90.1% higher; write p95: SQLite was 23.2% faster | +65.6% | -30.2% | -22.7% | +90.1% | -23.2% | 204.1 MB | 164.1 MB |
SQLite PR #378 vs SQLite rc3
Each row compares SQLite PR #378 against rc3 for the same workload and number of concurrent requests.
| Workload | Concurrent requests | Plain English | SQLite PR #378 flows/s | SQLite rc3 flows/s | SQLite PR #378 throughput vs SQLite rc3 | SQLite PR #378 avg flow | SQLite rc3 avg flow | SQLite PR #378 avg flow vs SQLite rc3 | SQLite PR #378 avg HTTP | SQLite rc3 avg HTTP | SQLite PR #378 avg HTTP vs SQLite rc3 | SQLite PR #378 p95 | SQLite rc3 p95 | SQLite PR #378 p95 vs SQLite rc3 | SQLite PR #378 write/s | SQLite rc3 write/s | SQLite PR #378 write throughput vs SQLite rc3 | SQLite PR #378 write p95 | SQLite rc3 write p95 | SQLite PR #378 write p95 vs SQLite rc3 | SQLite PR #378 failed | SQLite PR #378 SQLite locks |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| read-heavy | 4 | throughput: SQLite was 6.0% lower; p95 latency: SQLite was 8.8% slower | 38.35 | 40.79 | -6.0% | 104 ms | 98 ms | +6.3% | 104 ms | 98 ms | +6.3% | 189 ms | 174 ms | +8.8% | 0.00 | 0.00 | 0 | 0 | ||||
| read-heavy | 16 | throughput: about the same; p95 latency: SQLite was 4.3% faster | 37.70 | 37.51 | +0.5% | 423 ms | 424 ms | -0.4% | 422 ms | 424 ms | -0.4% | 775 ms | 810 ms | -4.3% | 0.00 | 0.00 | 0 | 0 | ||||
| read-heavy | 32 | throughput: SQLite was 6.3% lower; p95 latency: SQLite was 0.8% faster | 34.56 | 36.88 | -6.3% | 916 ms | 862 ms | +6.3% | 916 ms | 862 ms | +6.3% | 1,519 ms | 1,531 ms | -0.8% | 0.00 | 0.00 | 0 | 0 | ||||
| read-heavy | 64 | throughput: SQLite was 27.1% lower; p95 latency: SQLite was 39.4% slower | 23.51 | 32.24 | -27.1% | 2,652 ms | 1,947 ms | +36.2% | 2,652 ms | 1,947 ms | +36.2% | 4,551 ms | 3,264 ms | +39.4% | 0.00 | 0.00 | 0 | 0 | ||||
| balanced | 4 | throughput: SQLite was 11.4% lower; p95 latency: SQLite was 10.1% slower; write throughput: SQLite was 4.6% lower; write p95: SQLite was 5.2% slower | 15.94 | 17.99 | -11.4% | 248 ms | 221 ms | +11.9% | 119 ms | 110 ms | +9.0% | 728 ms | 661 ms | +10.1% | 3.79 | 3.97 | -4.6% | 758 ms | 720 ms | +5.2% | 0 | 0 |
| balanced | 16 | throughput: SQLite was 3.0% higher; p95 latency: SQLite was 11.8% faster; write throughput: SQLite was 4.5% higher; write p95: SQLite was 6.0% faster | 18.21 | 17.68 | +3.0% | 860 ms | 887 ms | -3.1% | 436 ms | 446 ms | -2.3% | 2,428 ms | 2,752 ms | -11.8% | 3.38 | 3.23 | +4.5% | 2,936 ms | 3,121 ms | -6.0% | 0 | 0 |
| balanced | 32 | throughput: SQLite was 1.7% higher; p95 latency: SQLite was 5.4% slower; write throughput: SQLite was 2.7% higher; write p95: SQLite was 3.3% slower | 19.19 | 18.86 | +1.7% | 1,610 ms | 1,642 ms | -2.0% | 831 ms | 845 ms | -1.7% | 4,608 ms | 4,371 ms | +5.4% | 3.82 | 3.72 | +2.7% | 4,938 ms | 4,781 ms | +3.3% | 0 | 0 |
| balanced | 64 | throughput: SQLite was 24.1% lower; p95 latency: SQLite was 29.8% slower; write throughput: SQLite was 25.8% lower; write p95: SQLite was 35.5% slower | 12.53 | 16.50 | -24.1% | 4,821 ms | 3,688 ms | +30.7% | 2,401 ms | 1,812 ms | +32.5% | 11,068 ms | 8,525 ms | +29.8% | 2.52 | 3.40 | -25.8% | 12,624 ms | 9,315 ms | +35.5% | 0 | 0 |
| write-heavy | 4 | throughput: SQLite was 2.4% lower; p95 latency: about the same; write throughput: SQLite was 1.3% lower; write p95: SQLite was 9.5% slower | 11.06 | 11.34 | -2.4% | 359 ms | 349 ms | +3.0% | 124 ms | 121 ms | +2.6% | 703 ms | 703 ms | -0.0% | 5.48 | 5.55 | -1.3% | 794 ms | 725 ms | +9.5% | 0 | 0 |
| write-heavy | 16 | throughput: SQLite was 1.7% higher; p95 latency: SQLite was 10.7% slower; write throughput: SQLite was 0.6% higher; write p95: SQLite was 9.5% slower | 11.24 | 11.04 | +1.7% | 1,403 ms | 1,410 ms | -0.5% | 476 ms | 478 ms | -0.4% | 3,042 ms | 2,747 ms | +10.7% | 5.55 | 5.52 | +0.6% | 3,189 ms | 2,912 ms | +9.5% | 0 | 0 |
| write-heavy | 32 | throughput: SQLite was 10.1% lower; p95 latency: SQLite was 9.1% slower; write throughput: SQLite was 8.5% lower; write p95: SQLite was 4.0% slower | 10.35 | 11.51 | -10.1% | 3,017 ms | 2,702 ms | +11.7% | 1,062 ms | 957 ms | +11.0% | 6,281 ms | 5,758 ms | +9.1% | 5.08 | 5.55 | -8.5% | 6,496 ms | 6,247 ms | +4.0% | 0 | 0 |
| write-heavy | 64 | throughput: SQLite was 20.6% lower; p95 latency: SQLite was 16.1% slower; write throughput: SQLite was 19.3% lower; write p95: SQLite was 13.8% slower | 8.33 | 10.49 | -20.6% | 7,236 ms | 5,841 ms | +23.9% | 2,491 ms | 2,012 ms | +23.8% | 13,971 ms | 12,037 ms | +16.1% | 4.13 | 5.12 | -19.3% | 14,529 ms | 12,767 ms | +13.8% | 0 | 0 |
Fixture And Host
WordPress 7.0; PHP 8.5.7 (cli) (built: Jun 2 2026 20:59:56) (NTS); nginx version: nginx/1.31.1
Fixture
- products: 2500
- categories: 100
- customers: 1000
- historical_orders: 10000
- product_meta_per_product: 16
- target_database_bytes: 5368709120
- filler_chunk_bytes: 262144
Run Integrity
- Cases: 144
- Failed flows: 361
- Write verification failures: 361
- HTTP 5xx: 0
- PHP fatals: 36
- SQLite lock errors: 3438
- Host: Apple M4 Max (Virtual), 4 CPUs
Absolute Results
read-heavy
Absolute medians across three repetitions.
| Variant | Concurrent requests | Flows/s | Avg flow | Avg HTTP request | p50 | p75 | p95 | p99 | Write/s | Write p95 | Failed | Verify fail | SQLite locks | Max WAL |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MariaDB baseline | 4 | 46.76 | 85 ms | 85 ms | 89 ms | 97 ms | 147 ms | 181 ms | 0.00 | 0 | 0 | 0 | ||
| SQLite rc3 | 4 | 40.79 | 98 ms | 98 ms | 94 ms | 116 ms | 174 ms | 233 ms | 0.00 | 0 | 0 | 0 | 4.0 MB | |
| SQLite stable | 4 | 40.42 | 99 ms | 99 ms | 96 ms | 118 ms | 179 ms | 223 ms | 0.00 | 0 | 0 | 58 | 4.0 MB | |
| SQLite PR #378 | 4 | 38.35 | 104 ms | 104 ms | 98 ms | 122 ms | 189 ms | 235 ms | 0.00 | 0 | 0 | 0 | 4.0 MB | |
| MariaDB baseline | 16 | 47.08 | 339 ms | 339 ms | 255 ms | 456 ms | 857 ms | 1,130 ms | 0.00 | 0 | 0 | 0 | ||
| SQLite rc3 | 16 | 37.51 | 424 ms | 424 ms | 420 ms | 501 ms | 810 ms | 994 ms | 0.00 | 0 | 0 | 0 | 4.3 MB | |
| SQLite stable | 16 | 37.63 | 423 ms | 423 ms | 428 ms | 497 ms | 807 ms | 1,033 ms | 0.00 | 0 | 0 | 36 | 4.2 MB | |
| SQLite PR #378 | 16 | 37.70 | 423 ms | 422 ms | 419 ms | 502 ms | 775 ms | 994 ms | 0.00 | 0 | 0 | 0 | 4.8 MB | |
| MariaDB baseline | 32 | 45.20 | 703 ms | 703 ms | 538 ms | 1,001 ms | 1,927 ms | 2,396 ms | 0.00 | 0 | 0 | 0 | ||
| SQLite rc3 | 32 | 36.88 | 862 ms | 862 ms | 862 ms | 1,012 ms | 1,531 ms | 1,835 ms | 0.00 | 0 | 0 | 0 | 4.9 MB | |
| SQLite stable | 32 | 36.85 | 860 ms | 860 ms | 859 ms | 1,007 ms | 1,515 ms | 1,868 ms | 0.00 | 0 | 0 | 50 | 6.1 MB | |
| SQLite PR #378 | 32 | 34.56 | 916 ms | 916 ms | 904 ms | 1,105 ms | 1,519 ms | 1,808 ms | 0.00 | 0 | 0 | 0 | 6.0 MB | |
| MariaDB baseline | 64 | 44.25 | 1,425 ms | 1,425 ms | 1,161 ms | 1,858 ms | 4,115 ms | 4,972 ms | 0.00 | 0 | 0 | 0 | ||
| SQLite rc3 | 64 | 32.24 | 1,947 ms | 1,947 ms | 1,881 ms | 2,372 ms | 3,264 ms | 4,193 ms | 0.00 | 0 | 0 | 0 | 12.7 MB | |
| SQLite stable | 64 | 35.30 | 1,786 ms | 1,786 ms | 1,745 ms | 2,164 ms | 3,076 ms | 3,954 ms | 0.00 | 0 | 0 | 90 | 6.6 MB | |
| SQLite PR #378 | 64 | 23.51 | 2,652 ms | 2,652 ms | 2,565 ms | 3,271 ms | 4,551 ms | 5,622 ms | 0.00 | 0 | 0 | 0 | 11.8 MB |
balanced
Absolute medians across three repetitions.
| Variant | Concurrent requests | Flows/s | Avg flow | Avg HTTP request | p50 | p75 | p95 | p99 | Write/s | Write p95 | Failed | Verify fail | SQLite locks | Max WAL |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MariaDB baseline | 4 | 16.94 | 235 ms | 113 ms | 110 ms | 255 ms | 673 ms | 780 ms | 4.06 | 777 ms | 0 | 0 | 0 | |
| SQLite rc3 | 4 | 17.99 | 221 ms | 110 ms | 122 ms | 202 ms | 661 ms | 726 ms | 3.97 | 720 ms | 0 | 0 | 0 | 6.5 MB |
| SQLite stable | 4 | 18.30 | 215 ms | 108 ms | 123 ms | 202 ms | 636 ms | 691 ms | 3.89 | 691 ms | 15 | 15 | 206 | 6.5 MB |
| SQLite PR #378 | 4 | 15.94 | 248 ms | 119 ms | 131 ms | 255 ms | 728 ms | 761 ms | 3.79 | 758 ms | 0 | 0 | 0 | 9.3 MB |
| MariaDB baseline | 16 | 23.16 | 679 ms | 339 ms | 209 ms | 483 ms | 2,726 ms | 3,099 ms | 4.62 | 3,100 ms | 0 | 0 | 0 | |
| SQLite rc3 | 16 | 17.68 | 887 ms | 446 ms | 584 ms | 850 ms | 2,752 ms | 3,120 ms | 3.23 | 3,121 ms | 0 | 0 | 0 | 80.7 MB |
| SQLite stable | 16 | 19.14 | 815 ms | 420 ms | 599 ms | 861 ms | 2,288 ms | 2,808 ms | 3.39 | 2,864 ms | 28 | 28 | 313 | 43.7 MB |
| SQLite PR #378 | 16 | 18.21 | 860 ms | 436 ms | 581 ms | 886 ms | 2,428 ms | 2,927 ms | 3.38 | 2,936 ms | 0 | 0 | 0 | 121.7 MB |
| MariaDB baseline | 32 | 25.00 | 1,248 ms | 628 ms | 374 ms | 891 ms | 5,053 ms | 5,426 ms | 5.15 | 5,420 ms | 0 | 0 | 0 | |
| SQLite rc3 | 32 | 18.86 | 1,642 ms | 845 ms | 1,153 ms | 1,912 ms | 4,371 ms | 4,772 ms | 3.72 | 4,781 ms | 0 | 0 | 0 | 120.9 MB |
| SQLite stable | 32 | 12.83 | 2,373 ms | 1,270 ms | 1,871 ms | 3,125 ms | 5,885 ms | 7,914 ms | 2.16 | 8,085 ms | 31 | 31 | 312 | 82.7 MB |
| SQLite PR #378 | 32 | 19.19 | 1,610 ms | 831 ms | 1,068 ms | 1,818 ms | 4,608 ms | 4,938 ms | 3.82 | 4,938 ms | 0 | 0 | 0 | 133.8 MB |
| MariaDB baseline | 64 | 24.48 | 2,473 ms | 1,215 ms | 961 ms | 2,012 ms | 9,864 ms | 10,585 ms | 5.02 | 10,563 ms | 0 | 0 | 0 | |
| SQLite rc3 | 64 | 16.50 | 3,688 ms | 1,812 ms | 2,943 ms | 4,637 ms | 8,525 ms | 9,317 ms | 3.40 | 9,315 ms | 0 | 0 | 0 | 90.3 MB |
| SQLite stable | 64 | 11.44 | 5,079 ms | 2,647 ms | 4,313 ms | 6,917 ms | 12,157 ms | 14,517 ms | 1.90 | 14,526 ms | 37 | 37 | 442 | 68.3 MB |
| SQLite PR #378 | 64 | 12.53 | 4,821 ms | 2,401 ms | 3,924 ms | 6,423 ms | 11,068 ms | 12,633 ms | 2.52 | 12,624 ms | 0 | 0 | 0 | 106.5 MB |
write-heavy
Absolute medians across three repetitions.
| Variant | Concurrent requests | Flows/s | Avg flow | Avg HTTP request | p50 | p75 | p95 | p99 | Write/s | Write p95 | Failed | Verify fail | SQLite locks | Max WAL |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MariaDB baseline | 4 | 10.03 | 396 ms | 138 ms | 449 ms | 628 ms | 761 ms | 979 ms | 5.10 | 818 ms | 0 | 0 | 0 | |
| SQLite rc3 | 4 | 11.34 | 349 ms | 121 ms | 224 ms | 550 ms | 703 ms | 812 ms | 5.55 | 725 ms | 0 | 0 | 0 | 9.6 MB |
| SQLite stable | 4 | 11.93 | 330 ms | 116 ms | 213 ms | 527 ms | 657 ms | 702 ms | 5.64 | 688 ms | 26 | 26 | 294 | 10.4 MB |
| SQLite PR #378 | 4 | 11.06 | 359 ms | 124 ms | 296 ms | 562 ms | 703 ms | 896 ms | 5.48 | 794 ms | 0 | 0 | 0 | 9.5 MB |
| MariaDB baseline | 16 | 13.08 | 1,197 ms | 406 ms | 1,328 ms | 2,113 ms | 2,489 ms | 3,015 ms | 6.57 | 2,870 ms | 0 | 0 | 0 | |
| SQLite rc3 | 16 | 11.04 | 1,410 ms | 478 ms | 1,153 ms | 2,253 ms | 2,747 ms | 3,130 ms | 5.52 | 2,912 ms | 0 | 0 | 0 | 116.0 MB |
| SQLite stable | 16 | 6.68 | 2,268 ms | 802 ms | 1,868 ms | 3,854 ms | 5,002 ms | 5,495 ms | 2.97 | 5,419 ms | 52 | 52 | 396 | 98.1 MB |
| SQLite PR #378 | 16 | 11.24 | 1,403 ms | 476 ms | 1,010 ms | 2,216 ms | 3,042 ms | 3,362 ms | 5.55 | 3,189 ms | 0 | 0 | 0 | 169.0 MB |
| MariaDB baseline | 32 | 15.29 | 2,046 ms | 715 ms | 1,836 ms | 3,693 ms | 4,299 ms | 4,750 ms | 7.60 | 4,580 ms | 0 | 0 | 0 | |
| SQLite rc3 | 32 | 11.51 | 2,702 ms | 957 ms | 2,068 ms | 4,058 ms | 5,758 ms | 6,538 ms | 5.55 | 6,247 ms | 0 | 0 | 0 | 208.0 MB |
| SQLite stable | 32 | 6.84 | 4,233 ms | 1,551 ms | 3,705 ms | 6,764 ms | 9,204 ms | 10,000 ms | 3.11 | 9,817 ms | 61 | 61 | 470 | 153.1 MB |
| SQLite PR #378 | 32 | 10.35 | 3,017 ms | 1,062 ms | 2,000 ms | 4,965 ms | 6,281 ms | 6,560 ms | 5.08 | 6,496 ms | 0 | 0 | 0 | 194.6 MB |
| MariaDB baseline | 64 | 15.50 | 3,928 ms | 1,355 ms | 2,631 ms | 7,254 ms | 8,599 ms | 9,104 ms | 7.58 | 9,023 ms | 0 | 0 | 0 | |
| SQLite rc3 | 64 | 10.49 | 5,841 ms | 2,012 ms | 4,484 ms | 9,082 ms | 12,037 ms | 13,508 ms | 5.12 | 12,767 ms | 0 | 0 | 0 | 204.1 MB |
| SQLite stable | 64 | 6.33 | 8,364 ms | 3,009 ms | 7,925 ms | 11,820 ms | 15,573 ms | 17,331 ms | 2.69 | 16,626 ms | 111 | 111 | 771 | 164.1 MB |
| SQLite PR #378 | 64 | 8.33 | 7,236 ms | 2,491 ms | 6,117 ms | 11,555 ms | 13,971 ms | 15,465 ms | 4.13 | 14,529 ms | 0 | 0 | 0 | 171.1 MB |
Worst Tail-Latency Cases
Sorted by measured p95 across all individual cases.
| Case | Variant | Workload | Concurrent requests | Flows/s | Avg flow | Avg HTTP request | p50 | p75 | p95 | p99 | Write/s | Write p95 | Failed |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| rep3_sqlite-plugin-2.2.23-control_write-heavy_c64 | SQLite stable | write-heavy | 64 | 5.61 | 9,609 ms | 3,474 ms | 7,925 ms | 15,636 ms | 19,540 ms | 20,877 ms | 2.46 | 36 | |
| rep2_sqlite-plugin-2.2.23-control_write-heavy_c64 | SQLite stable | write-heavy | 64 | 6.33 | 8,364 ms | 3,009 ms | 8,169 ms | 11,820 ms | 15,573 ms | 17,331 ms | 2.69 | 41 | |
| rep1_sqlite-driver-pr-378-performance_write-heavy_c64 | SQLite PR #378 | write-heavy | 64 | 8.33 | 7,236 ms | 2,491 ms | 6,117 ms | 11,555 ms | 14,123 ms | 15,775 ms | 4.13 | 0 | |
| rep2_sqlite-driver-pr-378-performance_write-heavy_c64 | SQLite PR #378 | write-heavy | 64 | 8.27 | 7,405 ms | 2,572 ms | 5,953 ms | 11,934 ms | 13,971 ms | 15,465 ms | 3.97 | 0 | |
| rep3_sqlite-plugin-2.2.23-control_balanced_c64 | SQLite stable | balanced | 64 | 9.34 | 6,004 ms | 2,976 ms | 5,329 ms | 8,130 ms | 13,247 ms | 15,539 ms | 1.86 | 9 | |
| rep3_sqlite-driver-pr-378-performance_write-heavy_c64 | SQLite PR #378 | write-heavy | 64 | 8.38 | 7,086 ms | 2,419 ms | 6,446 ms | 10,657 ms | 12,645 ms | 13,547 ms | 4.19 | 0 | |
| rep1_sqlite-driver-3.0.0-rc.3_write-heavy_c64 | SQLite rc3 | write-heavy | 64 | 9.63 | 6,313 ms | 2,146 ms | 5,036 ms | 10,082 ms | 12,543 ms | 13,984 ms | 4.83 | 0 | |
| rep2_sqlite-plugin-2.2.23-control_balanced_c64 | SQLite stable | balanced | 64 | 11.44 | 5,079 ms | 2,647 ms | 4,313 ms | 6,917 ms | 12,157 ms | 14,517 ms | 1.90 | 17 | |
| rep2_sqlite-driver-3.0.0-rc.3_write-heavy_c64 | SQLite rc3 | write-heavy | 64 | 10.49 | 5,841 ms | 2,012 ms | 4,484 ms | 9,082 ms | 12,037 ms | 13,508 ms | 5.12 | 0 | |
| rep3_sqlite-driver-3.0.0-rc.3_write-heavy_c64 | SQLite rc3 | write-heavy | 64 | 11.00 | 5,532 ms | 1,871 ms | 4,337 ms | 8,544 ms | 11,764 ms | 12,937 ms | 5.50 | 0 |