LocalStorage First Read Benchmark

A benchmark tool that measures the first localStorage access time in a fresh browser instance. Includes a web UI for manual testing and PowerShell scripts for automated multi-run comparisons across different DOM Storage backends.

Files

File	Description
index.html	Benchmark web page (manual UI + URL-driven automation)
benchmark.ps1	Automated single-origin cold-read benchmark
benchmark-warm-db.ps1	Automated cross-origin warm-DB benchmark

Manual Testing

1. Open the benchmark page

Use the deployed GitHub Pages site:

https://rahulsingh-msft.github.io/localstorage-getall-perf/

Or serve locally:

python -m http.server 9090

2. Launch the browser with the desired backend

SQLite backend:

msedge.exe --enable-features=DomStorageSqlite `
           --user-data-dir="$env:TEMP\edge-sqlite-test" `
           https://rahulsingh-msft.github.io/localstorage-getall-perf/

LevelDB backend (default):

msedge.exe --disable-features=DomStorageSqlite `
           --user-data-dir="$env:TEMP\edge-leveldb-test" `
           https://rahulsingh-msft.github.io/localstorage-getall-perf/

Use separate --user-data-dir paths so databases don't interfere.

3. Benchmark workflow

1. Populate — select entry count and value size, click Populate
2. Close the browser — ensures data is persisted to disk
3. Reopen the browser with the same --user-data-dir
4. Measure — click Measure to time the first localStorage read

Use the Delay before read option to let browser startup I/O settle.

Automated Testing

Cold-read benchmark (benchmark.ps1)

Measures the first localStorage read after a browser restart across multiple runs for both LevelDB and SQLite backends.

.\benchmark.ps1 -Runs 10 -Entries 10000 -ValueSize 100 -Delay 5000

Results are written to results.txt.

Warm-DB benchmark (benchmark-warm-db.ps1)

Isolates DB opening overhead from data retrieval cost. Populates two origins on separate ports, then measures:

• Tab 1 (origin A) — cold read (DB not yet open)
• Tab 2 (origin B) — warm-DB read (DB already open, data not cached)

The delta between the two reveals how much time is spent opening the database vs reading data.

.\benchmark-warm-db.ps1 -Runs 10 -Entries 10000 -Delay 5000

Results are written to results-warm-db.txt.

Requires Python for local HTTP servers (two ports for two origins).

Common parameters

Parameter	Default	Description
-Runs	10	Measurement cycles per backend
-Entries	10000	localStorage entries to populate
-ValueSize	100	Characters per value
-Delay	5000	Milliseconds to wait before measuring
-EdgePath	Edge SxS	Path to msedge.exe

URL parameters for automation

The page supports URL parameters for script-driven use:

Parameter	Example	Description
?populate=N&valueSize=V	?populate=10000&valueSize=100	Auto-populates localStorage, sets document.title to POPULATED:<count>
?auto&delay=Ms	?auto&delay=5000	Waits, then measures first read, sets document.title to RESULT:<ms>:<entries>

Notes

• Why restart? — Browsers cache localStorage in memory at multiple levels. These caches persist across tab opens/closes. Restarting guarantees a true first read from disk.
• Warm read detection — if you populate and measure without restarting, the UI flags the result as warm.
• Quota — localStorage allows ~5M characters per origin (~10 MB as UTF-16).