Page load metricsedit
Duration of the DNS query for the current page:
Making a connection to the server
Duration of the TCP connection to the server, including the TLS negotiation time for HTTPS pages:
Requesting and receiving the document
Overall response time of the server including the last byte:
Parsing the document, executing sync. scripts
HTML document parsing time, including synchronous Stylesheets and Script:
Fire "DOMContentLoaded" event
Triggered when the browser completes parsing the document. Helpful when there are multiple listeners, or logic
Fire "load" event
Trigged when the browser finishes loading its document and dependent resources:
Time to first byte (TTFB)
Duration from the browser making an HTTP request for the initial document to the first byte of the page being received. TTFB is available from the Navigation Timing API as the
reponseStarttimestamp, and captured as transaction mark
transaction.marks.agent.timeToFirstBytein the ElasticSearch document.
To capture the overall user experience of the page including all of the above information plus additional resource requests that might be
triggered during the execution of dependent resources, the
page-load transaction duration might not always reflect the
Load event of the browser and can extend beyond the event.
If you are interested in accurately measuring the duration of load event, the information can be extracted by using
Fire load event Span or from the Transaction marks available as
transaction.marks.agent.domComplete in the Elasticsearch document.
The page load transaction is measured relative to the
fetchStart timestamp from the Navigation Timing API.
To understand the performance characteristics of a web page beyond the page load and how the user perceives performance, the agent supports capturing these responsiveness metrics.
First contentful paint (FCP)
Focusses on the initial rendering and measures the time from when the page starts loading to when any part of the page’s content is displayed on the screen. The agent uses the Paint timing API available in the browser to capture the timing information. FCP is captured as transaction mark for
page-loadtransaction for all chromium-based browsers (Chrome >60, Edge >79, Opera >47, etc.).
Largest contentful paint (LCP)
A new page-load metric that denotes the time from when the page starts loading to when the critical element (largest text, image, video elements) is displayed on the screen. LCP is available in the browser through
LargestContentfulPaint API which relies on the draft Element Timing API. LCP is one of the core web vitals metrics and
available only in Chrome >77. Captured as transaction mark for
page-loadtransaction, maintain LCP within the first 2.5 seconds of page-load to provide a good user experience.
First input delay (FID)
FID quantifies the experience of the user when they interact with the page during the page load. It is measured as the time between when a user first interacts with your site (mouse clicks, taps, select dropdowns, etc.) to the time when the
browser can respond to that interaction. FID is one of the core web vitals metrics and available only in Chrome >85 via Event Timing API. FID is captured as
First Input Delayspan for
page-loadtransaction. Maintain FID below 100 milliseconds to provide a good user experience.
Total blocking time (TBT)
The sum of the blocking time (duration above 50 ms) for each long task that occurs between the First contentful paint and the time when the transaction is completed. Total blocking time is a
great companion metric for Time to interactive (TTI) which is lab metric and not available in the field through browser APIs. The agent captures TBT based on the number of long tasks that occurred during the page load lifecycle. Captured as
Total Blocking Timespan for
Cumulative layout shift (CLS)
- Metric that represents the cumulative score of all unexpected layout shifts that occur during the entire lifespan of the page. CLS is one of the core web vitals metrics and here is how the score is calculated by Chrome. Maintain a score of less than 0.1 to provide a good user experience.
A long task is any user activity or browser task that monopolize the UI thread for extended periods (greater than 50 milliseconds) and block other critical tasks (frame rate or input latency)
from being executed. The agent uses the Long Task API which is only available in chromium-based browsers (Chrome >58, Edge >79, Opera >69). Captured as
Longtask<name>span for all managed transactions. Learn more about long task spans, and how to interpret them.
- The User Timing spec exposes API for developers to add custom performance entries to their web application. These custom metrics allow users to measure the hot paths of the app code that helps to identify a good user experience. RUM agent records all of the performance.measure calls as spans by their measure name in the waterfall for all managed transactions.
The RUM agent automatically instruments click event listeners that are
registered by the application. The click events are captured as
transactions. However, to avoid sending too many
to the server, the agent discards transactions with no spans (e.g. no network activity). Furthermore,
if the click interaction results in route change, then a
transaction would be captured instead.
Single Page Applicationsedit
pushState events will be captured as transactions.
Most of these transactions can be enhanced by using framework specific integrations.
For all unsupported frameworks/libraries, you can instrument the application
by creating custom transactions and custom spans with the span API.
The agent supports integrations with certain frameworks.
To instrument custom metrics, like rendering time or mounting time of components on frameworks like React, Angular, Vue, etc., use the custom transactions API.
The following platforms are supported: