NETWORK SPEED TESTER

Resource waterfall

Inspect browser-recorded resource phases and overlap locally, from this page or a redacted capture of your own application.

Open resource waterfall — opens an ad-free tool page. Processing starts only after your explicit action.

Understand what your browser recorded

The resource waterfall displays the fetch intervals already recorded by the browser for a document. Start with Capture this page to inspect this diagnostic page, or import a read-only capture from your own application. Each resource is assigned its original capture-list index instead of displaying its URL. The timeline aligns starts and finishes against the same navigation-relative clock. It can show overlapping fetches and observable phases, but overlap does not establish which resource depends on another or which operation caused a delay. No additional test traffic is generated by capture.

Capture an application you control

Expand Analyse my own application and review the supplied console expression. Run it only in the developer console of a page you own or have permission to inspect. It reads existing resource and navigation entries and returns JSON text. Copy that resulting text, without the console display quotes, into the timing input. The expression removes names, URLs, origins, query strings, headers and server-timing descriptions before producing its output. The local parser treats pasted text only as data and does not execute it. Imported evidence is explicitly unverified; it cannot certify a production incident or prove that the supplied timestamps were honestly collected.

Read phases without counting the same time twice

The background bar spans the start of the observed fetch through the last received byte. Exposed DNS, connection, request-to-first-byte and response intervals are shown as overlays, with numeric values in a keyboard-accessible table. The connection interval includes any reported TLS negotiation; the separate TLS value is a subset, not an extra duration to add. Other work, including cache handling, redirects, scheduling or service-worker activity, may lie outside the displayed phases. The tool does not name those unallocated gaps as queue time. Request-to-first-byte is not server CPU time and cannot isolate wire latency.

Keep unavailable distinct from a measured zero

Cross-origin visibility controls, reuse and browser precision may hide or zero some fields. A missing protocol, HTTP status or detailed interval is shown as unavailable rather than guessed. Equal nonzero timing boundaries can produce an observed zero-duration phase. A reported transfer size of zero is retained as a browser value but does not alone prove a cache hit. The document fetch bar ends at responseEnd; navigation duration can instead end at loadEventEnd. DOM interactive, DOM content loaded and the end of the load event are separate milestones, not additional network-transfer durations.

Account for bounds and missing entries

A capture includes at most the first fifty retained resource entries plus one navigation entry. The full observed resource count is retained so the interface can report omissions. Input is bounded to 128 KiB, and the strict JSON reader also limits depth, values and string size. Invalid, unfinished or contradictory entries remain visible as counts or partial detail rather than being coerced into success. A browser may have discarded earlier entries before capture, and in-flight requests may not yet appear at all. This utility never clears or expands the performance buffer and cannot reconstruct its missing history.

Use local evidence, not a network verdict

All processing is local after the site assets have loaded. The runtime has an additive network-blocking policy; it does not upload captures, admit sessions, poll a server or store results. Explicit JSON export contains only generic identifiers and allowlisted values. Editing an import invalidates the old report; Clear, reload and page lifecycle transitions remove inputs and results. Compare captures under consistent foreground-browser conditions. These observations do not identify an ISP, router or Wi-Fi fault, verify ordinary CDN cache correctness, or reconstruct a full request dependency graph. The asset-integrity tool provides a separate byte-and-digest consistency check.

Method references and related tools

W3C Resource Timing defines resource fetch fields and timing visibility. Navigation Timing Level 2 defines document timing and lifecycle milestones. These evolving specifications do not certify this diagnostic.

Asset integrity · All tools · Overall measurement methodology