Performance measurement APIs | Node.js v17.9.1 Documentation (original) (raw)

Source Code: lib/perf_hooks.js

This module provides an implementation of a subset of the W3CWeb Performance APIs as well as additional APIs for Node.js-specific performance measurements.

Node.js supports the following Web Performance APIs:

const { PerformanceObserver, performance } = require('perf_hooks');

const obs = new PerformanceObserver((items) => {
  console.log(items.getEntries()[0].duration);
  performance.clearMarks();
});
obs.observe({ type: 'measure' });
performance.measure('Start to Now');

performance.mark('A');
doSomeLongRunningProcess(() => {
  performance.measure('A to Now', 'A');

  performance.mark('B');
  performance.measure('A to B', 'A', 'B');
});

perf_hooks.performance#

Added in: v8.5.0

An object that can be used to collect performance metrics from the current Node.js instance. It is similar to window.performance in browsers.

performance.clearMarks([name])#

Added in: v8.5.0

If name is not provided, removes all PerformanceMark objects from the Performance Timeline. If name is provided, removes only the named mark.

performance.clearMeasures([name])#

Added in: v16.7.0

If name is not provided, removes all PerformanceMeasure objects from the Performance Timeline. If name is provided, removes only the named mark.

performance.eventLoopUtilization([utilization1[, utilization2]])#

Added in: v14.10.0, v12.19.0

The eventLoopUtilization() method returns an object that contains the cumulative duration of time the event loop has been both idle and active as a high resolution milliseconds timer. The utilization value is the calculated Event Loop Utilization (ELU).

If bootstrapping has not yet finished on the main thread the properties have the value of 0. The ELU is immediately available on Worker threads since bootstrap happens within the event loop.

Both utilization1 and utilization2 are optional parameters.

If utilization1 is passed, then the delta between the current call's activeand idle times, as well as the corresponding utilization value are calculated and returned (similar to process.hrtime()).

If utilization1 and utilization2 are both passed, then the delta is calculated between the two arguments. This is a convenience option because, unlike process.hrtime(), calculating the ELU is more complex than a single subtraction.

ELU is similar to CPU utilization, except that it only measures event loop statistics and not CPU usage. It represents the percentage of time the event loop has spent outside the event loop's event provider (e.g. epoll_wait). No other CPU idle time is taken into consideration. The following is an example of how a mostly idle process will have a high ELU.

'use strict';
const { eventLoopUtilization } = require('perf_hooks').performance;
const { spawnSync } = require('child_process');

setImmediate(() => {
  const elu = eventLoopUtilization();
  spawnSync('sleep', ['5']);
  console.log(eventLoopUtilization(elu).utilization);
});

Although the CPU is mostly idle while running this script, the value ofutilization is 1. This is because the call tochild_process.spawnSync() blocks the event loop from proceeding.

Passing in a user-defined object instead of the result of a previous call toeventLoopUtilization() will lead to undefined behavior. The return values are not guaranteed to reflect any correct state of the event loop.

performance.getEntries()#

Added in: v16.7.0

Returns a list of PerformanceEntry objects in chronological order with respect to performanceEntry.startTime. If you are only interested in performance entries of certain types or that have certain names, seeperformance.getEntriesByType() and performance.getEntriesByName().

performance.getEntriesByName(name[, type])#

Added in: v16.7.0

Returns a list of PerformanceEntry objects in chronological order with respect to performanceEntry.startTime whose performanceEntry.name is equal to name, and optionally, whose performanceEntry.entryType is equal totype.

performance.getEntriesByType(type)#

Added in: v16.7.0

Returns a list of PerformanceEntry objects in chronological order with respect to performanceEntry.startTime whose performanceEntry.entryTypeis equal to type.

performance.mark([name[, options]])#

Creates a new PerformanceMark entry in the Performance Timeline. APerformanceMark is a subclass of PerformanceEntry whoseperformanceEntry.entryType is always 'mark', and whoseperformanceEntry.duration is always 0. Performance marks are used to mark specific significant moments in the Performance Timeline.

The created PerformanceMark entry is put in the global Performance Timeline and can be queried with performance.getEntries,performance.getEntriesByName, and performance.getEntriesByType. When the observation is performed, the entries should be cleared from the global Performance Timeline manually with performance.clearMarks.

performance.measure(name[, startMarkOrOptions[, endMark]])#

Creates a new PerformanceMeasure entry in the Performance Timeline. APerformanceMeasure is a subclass of PerformanceEntry whoseperformanceEntry.entryType is always 'measure', and whoseperformanceEntry.duration measures the number of milliseconds elapsed sincestartMark and endMark.

The startMark argument may identify any existing PerformanceMark in the Performance Timeline, or may identify any of the timestamp properties provided by the PerformanceNodeTiming class. If the named startMark does not exist, an error is thrown.

The optional endMark argument must identify any existing PerformanceMarkin the Performance Timeline or any of the timestamp properties provided by thePerformanceNodeTiming class. endMark will be performance.now()if no parameter is passed, otherwise if the named endMark does not exist, an error will be thrown.

The created PerformanceMeasure entry is put in the global Performance Timeline and can be queried with performance.getEntries,performance.getEntriesByName, and performance.getEntriesByType. When the observation is performed, the entries should be cleared from the global Performance Timeline manually with performance.clearMeasures.

performance.nodeTiming#

Added in: v8.5.0

This property is an extension by Node.js. It is not available in Web browsers.

An instance of the PerformanceNodeTiming class that provides performance metrics for specific Node.js operational milestones.

performance.now()#

Added in: v8.5.0

Returns the current high resolution millisecond timestamp, where 0 represents the start of the current node process.

performance.timeOrigin#

Added in: v8.5.0

The timeOrigin specifies the high resolution millisecond timestamp at which the current node process began, measured in Unix time.

performance.timerify(fn[, options])#

This property is an extension by Node.js. It is not available in Web browsers.

Wraps a function within a new function that measures the running time of the wrapped function. A PerformanceObserver must be subscribed to the 'function'event type in order for the timing details to be accessed.

const {
  performance,
  PerformanceObserver
} = require('perf_hooks');

function someFunction() {
  console.log('hello world');
}

const wrapped = performance.timerify(someFunction);

const obs = new PerformanceObserver((list) => {
  console.log(list.getEntries()[0].duration);

  performance.clearMarks();
  performance.clearMeasures();
  obs.disconnect();
});
obs.observe({ entryTypes: ['function'] });

// A performance timeline entry will be created
wrapped();

If the wrapped function returns a promise, a finally handler will be attached to the promise and the duration will be reported once the finally handler is invoked.

performance.toJSON()#

Added in: v16.1.0

An object which is JSON representation of the performance object. It is similar to window.performance.toJSON in browsers.

Class: PerformanceEntry#

Added in: v8.5.0

performanceEntry.detail#

Added in: v16.0.0

Additional detail specific to the entryType.

performanceEntry.duration#

Added in: v8.5.0

The total number of milliseconds elapsed for this entry. This value will not be meaningful for all Performance Entry types.

performanceEntry.entryType#

Added in: v8.5.0

The type of the performance entry. It may be one of:

performanceEntry.flags#

This property is an extension by Node.js. It is not available in Web browsers.

When performanceEntry.entryType is equal to 'gc', the performance.flagsproperty contains additional information about garbage collection operation. The value may be one of:

performanceEntry.name#

Added in: v8.5.0

The name of the performance entry.

performanceEntry.kind#

This property is an extension by Node.js. It is not available in Web browsers.

When performanceEntry.entryType is equal to 'gc', the performance.kindproperty identifies the type of garbage collection operation that occurred. The value may be one of:

performanceEntry.startTime#

Added in: v8.5.0

The high resolution millisecond timestamp marking the starting time of the Performance Entry.

Garbage Collection ('gc') Details#

When performanceEntry.type is equal to 'gc', the performanceEntry.detailproperty will be an with two properties:

HTTP/2 ('http2') Details#

When performanceEntry.type is equal to 'http2', theperformanceEntry.detail property will be an containing additional performance information.

If performanceEntry.name is equal to Http2Stream, the detailwill contain the following properties:

If performanceEntry.name is equal to Http2Session, the detail will contain the following properties:

Timerify ('function') Details#

When performanceEntry.type is equal to 'function', theperformanceEntry.detail property will be an listing the input arguments to the timed function.

Net ('net') Details#

When performanceEntry.type is equal to 'net', theperformanceEntry.detail property will be an containing additional information.

If performanceEntry.name is equal to connect, the detailwill contain the following properties: host, port.

DNS ('dns') Details#

When performanceEntry.type is equal to 'dns', theperformanceEntry.detail property will be an containing additional information.

If performanceEntry.name is equal to lookup, the detailwill contain the following properties: hostname, family, hints, verbatim.

If performanceEntry.name is equal to lookupService, the detail will contain the following properties: host, port.

If performanceEntry.name is equal to queryxxx or getHostByAddr, the detail will contain the following properties: host, ttl.

Class: PerformanceNodeTiming#

Added in: v8.5.0

This property is an extension by Node.js. It is not available in Web browsers.

Provides timing details for Node.js itself. The constructor of this class is not exposed to users.

performanceNodeTiming.bootstrapComplete#

Added in: v8.5.0

The high resolution millisecond timestamp at which the Node.js process completed bootstrapping. If bootstrapping has not yet finished, the property has the value of -1.

performanceNodeTiming.environment#

Added in: v8.5.0

The high resolution millisecond timestamp at which the Node.js environment was initialized.

performanceNodeTiming.idleTime#

Added in: v14.10.0, v12.19.0

The high resolution millisecond timestamp of the amount of time the event loop has been idle within the event loop's event provider (e.g. epoll_wait). This does not take CPU usage into consideration. If the event loop has not yet started (e.g., in the first tick of the main script), the property has the value of 0.

performanceNodeTiming.loopExit#

Added in: v8.5.0

The high resolution millisecond timestamp at which the Node.js event loop exited. If the event loop has not yet exited, the property has the value of -1. It can only have a value of not -1 in a handler of the 'exit' event.

performanceNodeTiming.loopStart#

Added in: v8.5.0

The high resolution millisecond timestamp at which the Node.js event loop started. If the event loop has not yet started (e.g., in the first tick of the main script), the property has the value of -1.

performanceNodeTiming.nodeStart#

Added in: v8.5.0

The high resolution millisecond timestamp at which the Node.js process was initialized.

performanceNodeTiming.v8Start#

Added in: v8.5.0

The high resolution millisecond timestamp at which the V8 platform was initialized.

Class: perf_hooks.PerformanceObserver#

new PerformanceObserver(callback)#

Added in: v8.5.0

PerformanceObserver objects provide notifications when newPerformanceEntry instances have been added to the Performance Timeline.

const {
  performance,
  PerformanceObserver
} = require('perf_hooks');

const obs = new PerformanceObserver((list, observer) => {
  console.log(list.getEntries());

  performance.clearMarks();
  performance.clearMeasures();
  observer.disconnect();
});
obs.observe({ entryTypes: ['mark'], buffered: true });

performance.mark('test');

Because PerformanceObserver instances introduce their own additional performance overhead, instances should not be left subscribed to notifications indefinitely. Users should disconnect observers as soon as they are no longer needed.

The callback is invoked when a PerformanceObserver is notified about new PerformanceEntry instances. The callback receives aPerformanceObserverEntryList instance and a reference to thePerformanceObserver.

performanceObserver.disconnect()#

Added in: v8.5.0

Disconnects the PerformanceObserver instance from all notifications.

performanceObserver.observe(options)#

Subscribes the instance to notifications of new instances identified either by options.entryTypesor options.type:

const {
  performance,
  PerformanceObserver
} = require('perf_hooks');

const obs = new PerformanceObserver((list, observer) => {
  // Called once asynchronously. `list` contains three items.
});
obs.observe({ type: 'mark' });

for (let n = 0; n < 3; n++)
  performance.mark(`test${n}`);

Class: PerformanceObserverEntryList#

Added in: v8.5.0

The PerformanceObserverEntryList class is used to provide access to thePerformanceEntry instances passed to a PerformanceObserver. The constructor of this class is not exposed to users.

performanceObserverEntryList.getEntries()#

Added in: v8.5.0

Returns a list of PerformanceEntry objects in chronological order with respect to performanceEntry.startTime.

const {
  performance,
  PerformanceObserver
} = require('perf_hooks');

const obs = new PerformanceObserver((perfObserverList, observer) => {
  console.log(perfObserverList.getEntries());
  /**
   * [
   *   PerformanceEntry {
   *     name: 'test',
   *     entryType: 'mark',
   *     startTime: 81.465639,
   *     duration: 0
   *   },
   *   PerformanceEntry {
   *     name: 'meow',
   *     entryType: 'mark',
   *     startTime: 81.860064,
   *     duration: 0
   *   }
   * ]
   */

  performance.clearMarks();
  performance.clearMeasures();
  observer.disconnect();
});
obs.observe({ type: 'mark' });

performance.mark('test');
performance.mark('meow');

performanceObserverEntryList.getEntriesByName(name[, type])#

Added in: v8.5.0

Returns a list of PerformanceEntry objects in chronological order with respect to performanceEntry.startTime whose performanceEntry.name is equal to name, and optionally, whose performanceEntry.entryType is equal totype.

const {
  performance,
  PerformanceObserver
} = require('perf_hooks');

const obs = new PerformanceObserver((perfObserverList, observer) => {
  console.log(perfObserverList.getEntriesByName('meow'));
  /**
   * [
   *   PerformanceEntry {
   *     name: 'meow',
   *     entryType: 'mark',
   *     startTime: 98.545991,
   *     duration: 0
   *   }
   * ]
   */
  console.log(perfObserverList.getEntriesByName('nope')); // []

  console.log(perfObserverList.getEntriesByName('test', 'mark'));
  /**
   * [
   *   PerformanceEntry {
   *     name: 'test',
   *     entryType: 'mark',
   *     startTime: 63.518931,
   *     duration: 0
   *   }
   * ]
   */
  console.log(perfObserverList.getEntriesByName('test', 'measure')); // []

  performance.clearMarks();
  performance.clearMeasures();
  observer.disconnect();
});
obs.observe({ entryTypes: ['mark', 'measure'] });

performance.mark('test');
performance.mark('meow');

performanceObserverEntryList.getEntriesByType(type)#

Added in: v8.5.0

Returns a list of PerformanceEntry objects in chronological order with respect to performanceEntry.startTime whose performanceEntry.entryTypeis equal to type.

const {
  performance,
  PerformanceObserver
} = require('perf_hooks');

const obs = new PerformanceObserver((perfObserverList, observer) => {
  console.log(perfObserverList.getEntriesByType('mark'));
  /**
   * [
   *   PerformanceEntry {
   *     name: 'test',
   *     entryType: 'mark',
   *     startTime: 55.897834,
   *     duration: 0
   *   },
   *   PerformanceEntry {
   *     name: 'meow',
   *     entryType: 'mark',
   *     startTime: 56.350146,
   *     duration: 0
   *   }
   * ]
   */
  performance.clearMarks();
  performance.clearMeasures();
  observer.disconnect();
});
obs.observe({ type: 'mark' });

performance.mark('test');
performance.mark('meow');

perf_hooks.createHistogram([options])#

Added in: v15.9.0, v14.18.0

Returns a .

perf_hooks.monitorEventLoopDelay([options])#

Added in: v11.10.0

This property is an extension by Node.js. It is not available in Web browsers.

Creates an IntervalHistogram object that samples and reports the event loop delay over time. The delays will be reported in nanoseconds.

Using a timer to detect approximate event loop delay works because the execution of timers is tied specifically to the lifecycle of the libuv event loop. That is, a delay in the loop will cause a delay in the execution of the timer, and those delays are specifically what this API is intended to detect.

const { monitorEventLoopDelay } = require('perf_hooks');
const h = monitorEventLoopDelay({ resolution: 20 });
h.enable();
// Do something.
h.disable();
console.log(h.min);
console.log(h.max);
console.log(h.mean);
console.log(h.stddev);
console.log(h.percentiles);
console.log(h.percentile(50));
console.log(h.percentile(99));

Class: Histogram#

Added in: v11.10.0

histogram.count#

Added in: v17.4.0

The number of samples recorded by the histogram.

histogram.countBigInt#

Added in: v17.4.0

The number of samples recorded by the histogram.

histogram.exceeds#

Added in: v11.10.0

The number of times the event loop delay exceeded the maximum 1 hour event loop delay threshold.

histogram.exceedsBigInt#

Added in: v17.4.0

The number of times the event loop delay exceeded the maximum 1 hour event loop delay threshold.

histogram.max#

Added in: v11.10.0

The maximum recorded event loop delay.

histogram.maxBigInt#

Added in: v17.4.0

The maximum recorded event loop delay.

histogram.mean#

Added in: v11.10.0

The mean of the recorded event loop delays.

histogram.min#

Added in: v11.10.0

The minimum recorded event loop delay.

histogram.minBigInt#

Added in: v17.4.0

The minimum recorded event loop delay.

histogram.percentile(percentile)#

Added in: v11.10.0

Returns the value at the given percentile.

histogram.percentileBigInt(percentile)#

Added in: v17.4.0

Returns the value at the given percentile.

histogram.percentiles#

Added in: v11.10.0

Returns a Map object detailing the accumulated percentile distribution.

histogram.percentilesBigInt#

Added in: v17.4.0

Returns a Map object detailing the accumulated percentile distribution.

histogram.reset()#

Added in: v11.10.0

Resets the collected histogram data.

histogram.stddev#

Added in: v11.10.0

The standard deviation of the recorded event loop delays.

Class: IntervalHistogram extends Histogram#

A Histogram that is periodically updated on a given interval.

histogram.disable()#

Added in: v11.10.0

Disables the update interval timer. Returns true if the timer was stopped, false if it was already stopped.

histogram.enable()#

Added in: v11.10.0

Enables the update interval timer. Returns true if the timer was started, false if it was already started.

Cloning an IntervalHistogram#

instances can be cloned via . On the receiving end, the histogram is cloned as a plain object that does not implement the enable() and disable() methods.

Class: RecordableHistogram extends Histogram#

Added in: v15.9.0, v14.18.0

histogram.add(other)#

Added in: v17.4.0

Adds the values from other to this histogram.

histogram.record(val)#

Added in: v15.9.0, v14.18.0

histogram.recordDelta()#

Added in: v15.9.0, v14.18.0

Calculates the amount of time (in nanoseconds) that has passed since the previous call to recordDelta() and records that amount in the histogram.

Examples#

Measuring the duration of async operations#

The following example uses the Async Hooks and Performance APIs to measure the actual duration of a Timeout operation (including the amount of time it took to execute the callback).

'use strict';
const async_hooks = require('async_hooks');
const {
  performance,
  PerformanceObserver
} = require('perf_hooks');

const set = new Set();
const hook = async_hooks.createHook({
  init(id, type) {
    if (type === 'Timeout') {
      performance.mark(`Timeout-${id}-Init`);
      set.add(id);
    }
  },
  destroy(id) {
    if (set.has(id)) {
      set.delete(id);
      performance.mark(`Timeout-${id}-Destroy`);
      performance.measure(`Timeout-${id}`,
                          `Timeout-${id}-Init`,
                          `Timeout-${id}-Destroy`);
    }
  }
});
hook.enable();

const obs = new PerformanceObserver((list, observer) => {
  console.log(list.getEntries()[0]);
  performance.clearMarks();
  performance.clearMeasures();
  observer.disconnect();
});
obs.observe({ entryTypes: ['measure'], buffered: true });

setTimeout(() => {}, 1000);

Measuring how long it takes to load dependencies#

The following example measures the duration of require() operations to load dependencies:

'use strict';
const {
  performance,
  PerformanceObserver
} = require('perf_hooks');
const mod = require('module');

// Monkey patch the require function
mod.Module.prototype.require =
  performance.timerify(mod.Module.prototype.require);
require = performance.timerify(require);

// Activate the observer
const obs = new PerformanceObserver((list) => {
  const entries = list.getEntries();
  entries.forEach((entry) => {
    console.log(`require('${entry[0]}')`, entry.duration);
  });
  performance.clearMarks();
  performance.clearMeasures();
  obs.disconnect();
});
obs.observe({ entryTypes: ['function'], buffered: true });

require('some-module');

Measuring how long one HTTP round-trip takes#

The following example is used to trace the time spent by HTTP client (OutgoingMessage) and HTTP request (IncomingMessage). For HTTP client, it means the time interval between starting the request and receiving the response, and for HTTP request, it means the time interval between receiving the request and sending the response:

'use strict';
const { PerformanceObserver } = require('perf_hooks');
const http = require('http');

const obs = new PerformanceObserver((items) => {
  items.getEntries().forEach((item) => {
    console.log(item);
  });
});

obs.observe({ entryTypes: ['http'] });

const PORT = 8080;

http.createServer((req, res) => {
  res.end('ok');
}).listen(PORT, () => {
  http.get(`http://127.0.0.1:${PORT}`);
});

Measuring how long the net.connect (only for TCP) takes when the connection is successful#

'use strict';
const { PerformanceObserver } = require('perf_hooks');
const net = require('net');
const obs = new PerformanceObserver((items) => {
  items.getEntries().forEach((item) => {
    console.log(item);
  });
});
obs.observe({ entryTypes: ['net'] });
const PORT = 8080;
net.createServer((socket) => {
  socket.destroy();
}).listen(PORT, () => {
  net.connect(PORT);
});

Measuring how long the DNS takes when the request is successful#

'use strict';
const { PerformanceObserver } = require('perf_hooks');
const dns = require('dns');
const obs = new PerformanceObserver((items) => {
  items.getEntries().forEach((item) => {
    console.log(item);
  });
});
obs.observe({ entryTypes: ['dns'] });
dns.lookup('localhost', () => {});
dns.promises.resolve('localhost');