The year is 2023 and we’ve released Node.js v20. It’s a significantaccomplishment, and this article aims to use scientific numbers to assess the state of Node.js’ performance.
All the benchmark results contain a reproducible example and hardware details. To reduce the noise for regular readers,the reproducible steps will be collapsed at the beginning of all sections.
This article aims to provide a comparative analysis of different versions of Node.js. It highlights the improvements andsetbacks and provides insights into the reasons behind those changes, without drawing any comparisons with otherJavaScript runtimes.
To conduct this experiment, we utilized Node.js versions 16.20.0, 18.16.0, and 20.0.0, and divided thebenchmark suites into three distinct groups:
- Node.js Internal Benchmark
Given the significant size and time-consuming nature of the Node.js benchmark suite, I have selectively chosenbenchmarks that, in my opinion, have a greater impact on Node.js developers and configurations, such as reading a filewith 16 MB using fs.readfile
. These benchmarks are grouped by modules, such as fs
and streams
. For additionaldetails on the Node.js benchmark suite, please refer to theNode.js source code.
I maintain a repository called nodejs-bench-operations
thatincludes benchmark operations for all major versions of Node.js, as well as the last three releases of each versionline. This allows for easy comparison of results between different versions, such as Node.js v16.20.0 and v18.16.0, orv19.8.0 and v19.9.0, with the objective of identifying regressions in the Node.js codebase. If you are interested inNode.js comparisons, following this repository might be beneficial (and don’t forget to give it a star if you find ithelpful).
- HTTP Servers (Frameworks)
This practical HTTP benchmark sends a significant number of requests to various routes, returning JSON, plain text, anderrors, taking express
and fastify
as references. The primary objective is to determine if the results obtained fromthe Node.js Internal Benchmark and nodejs-bench-operations areapplicable to common HTTP applications.
💡 UPDATE: Due to the extensive content covered in this article, the third and final step will be shared in asubsequent article. To stay updated and receive notifications, I encourage you to follow me onTwitter/LinkedIn.
Environment
To perform this benchmark, an AWS Dedicated Host was used with thefollowing computing-optimized instance:
- c6i.xlarge (Ice Lake) 3,5 GHz - Computing Optimized
- 4 vCPUs
- 8 GB Mem
- Canonical, Ubuntu, 22.04 LTS, amd64 jammy
- 1GiB SSD Volume Type
Node.js Internal Benchmark
The following modules/namespaces were selected in this benchmark:
fs
- Node.js file systemevents
- Node.js event classesEventEmitter
/EventTarget
http
- Node.js HTTP server + parsermisc
- Node.js startup time usingchild_processes
andworker_threads
+trace_events
module
- Node.jsmodule.require
streams
- Node.js streams creation, destroy, readable and moreurl
- Node.js URL parserbuffers
- Node.js Buffer operationsutil
- Node.js text encoder/decoder
And the configurations used are available atRafaelGSS/node#state-of-nodejs and all the results werepublished in the main repository:State of Node.js Performance 2023.
Node.js benchmark approach
Before presenting the results, it is crucial to explain the statistical approach used to determine the confidence of thebenchmark results. This method has been explained in detail in a previous blog post, which you can refer to here:Preparing and Evaluating Benchmarks.
To compare the impact of a new Node.js version, we ran each benchmark multiple times (30) on each configuration and onNode.js 16, 18, and 20. When the output is shown as a table, there are two columns that require careful attention:
- improvement - the percentage of improvement relative to the new version
- confidence - tells us if there is enough statistical evidence to validate the improvement
For example, consider the following table results:
confidence improvement accuracy (*) (**) (***)fs/readfile.js concurrent=1 len=16777216 encoding='ascii' duration=5 *** 67.59 % ±3.80% ±5.12% ±6.79%fs/readfile.js concurrent=1 len=16777216 encoding='utf-8' duration=5 *** 11.97 % ±1.09% ±1.46% ±1.93%fs/writefile-promises.js concurrent=1 size=1024 encodingType='utf' duration=5 0.36 % ±0.56% ±0.75% ±0.97%Be aware that when doing many comparisons the risk of a false-positive result increases.In this case, there are 10 comparisons, you can thus expect the following amount of false-positive results: 0.50 false positives, when considering a 5% risk acceptance (*, **, ***), 0.10 false positives, when considering a 1% risk acceptance (**, ***), 0.01 false positives, when considering a 0.1% risk acceptance (***)
There is a risk of 0.1% that fs.readfile
didn’t improve from Node.js 16 to Node.js 18 (confidence ***). Hence, weare pretty confident with the results. The table structure can be read as:
fs/readfile.js
- benchmark fileconcurrent=1 len=16777216 encoding='ascii' duration=5
- benchmark options. Each benchmark file can have manyoptions, in this case, it’s reading 1 concurrent file with 16777216 bytes during 5 seconds usingASCII as the encoding method.
For the statistically minded, the script performsan independent/unpaired 2-group t-test,with the null hypothesis that the performance is the same for both versions. The confidence field will show a star ifthe p-value is less than
0.05
. —Writing and Running benchmarks
Benchmark Setup
- Clone the fork Node.js repo
- Checkout
state-of-nodejs
branch - Create Node.js 16, 18, and 20 binaries
- Run the
benchmark.sh
script
#1git clone [emailprotected]:RafaelGSS/node.git#2cd node && git checkout state-of-nodejs#3nvm install v20.0.0cp $(which node) ./node20nvm install v18.16.0cp $(which node) ./node18nvm install v16.20.0cp $(which node) ./node16#4./benchmark.sh
File System
When upgrading Node.js from 16 to 18, an improvement of 67% was observed when using fs.readfile
API with an ascii
encoding and 12% roughly when using utf-8
.
The benchmark results showed that there was an improvement of about 67% in the fs.readfile
API with an ascii
encoding and roughly 12% when using utf-8
when upgrading Node.js from version 16 to 18. The file utilized for thebenchmark was created using the following code snippet:
const data = Buffer.alloc(16 * 1024 * 1024, 'x');fs.writeFileSync(filename, data);
However, there was a regression when using fs.readfile
with ascii
on Node.js 20 of 27%. This regression has beenreported to the Node.js Performance team, and it is expected to be fixed. On the other hand, fs.opendir
,fs.realpath
, and fs.readdir
showed improvement from Node.js 18 to Node.js 20. The comparison between Node.js 18 and20 can be seen in the benchmark result below:
confidence improvement accuracy (*) (**) (***)fs/bench-opendir.js bufferSize=1024 mode='async' dir='test/parallel' n=100 *** 3.48 % ±0.22% ±0.30% ±0.39%fs/bench-opendir.js bufferSize=32 mode='async' dir='test/parallel' n=100 *** 7.86 % ±0.29% ±0.39% ±0.50%fs/bench-readdir.js withFileTypes='false' dir='test/parallel' n=10 *** 8.69 % ±0.22% ±0.30% ±0.39%fs/bench-realpath.js pathType='relative' n=10000 *** 5.13 % ±0.97% ±1.29% ±1.69%fs/readfile.js concurrent=1 len=16777216 encoding='ascii' duration=5 *** -27.30 % ±4.27% ±5.75% ±7.63%fs/readfile.js concurrent=1 len=16777216 encoding='utf-8' duration=5 *** 3.25 % ±0.61% ±0.81% ±1.06% 0.10 false positives, when considering a 5% risk acceptance (*, **, ***), 0.02 false positives, when considering a 1% risk acceptance (**, ***), 0.00 false positives, when considering a 0.1% risk acceptance (***)
If you are using Node.js 16, you can use the following comparison between Node.js 16 and Node.js 20
confidence improvement accuracy (*) (**) (***)fs/bench-opendir.js bufferSize=1024 mode='async' dir='test/parallel' n=100 *** 2.79 % ±0.26% ±0.35% ±0.46%fs/bench-opendir.js bufferSize=32 mode='async' dir='test/parallel' n=100 *** 5.41 % ±0.27% ±0.35% ±0.46%fs/bench-readdir.js withFileTypes='false' dir='test/parallel' n=10 *** 2.19 % ±0.26% ±0.35% ±0.45%fs/bench-realpath.js pathType='relative' n=10000 *** 6.86 % ±0.94% ±1.26% ±1.64%fs/readfile.js concurrent=1 len=16777216 encoding='ascii' duration=5 *** 21.96 % ±7.96% ±10.63% ±13.92%fs/readfile.js concurrent=1 len=16777216 encoding='utf-8' duration=5 *** 15.55 % ±1.09% ±1.46% ±1.92%
Events
The EventTarget
class showed the most significant improvement on the events side. The benchmark involved dispatching amillion events using EventTarget.prototype.dispatchEvent(new Event('foo'))
.
Upgrading from Node.js 16 to Node.js 18 can deliver an improvement of nearly 15% in event dispatching performance.But the real jump comes when upgrading from Node.js 18 to Node.js 20, which can yield a performance improvement of up to200% when there is only a single listener.
The EventTarget
class is a crucial component of the Web API and is utilized in various parent features such asAbortSignal
and worker_threads
. As a result, optimizations made to this class can potentially impact the performanceof these features, including fetch
and AbortController
. Additionally, the EventEmitter.prototype.emit
API also sawa notable improvement of approximately 11.5% when comparing Node.js 16 to Node.js 20. A comprehensive comparison isprovided below for your reference:
confidence improvement accuracy (*) (**) (***)events/ee-emit.js listeners=5 argc=2 n=2000000 *** 11.49 % ±1.37% ±1.83% ±2.38%events/ee-once.js argc=0 n=20000000 *** -4.35 % ±0.47% ±0.62% ±0.81%events/eventtarget-add-remove.js nListener=10 n=1000000 *** 3.80 % ±0.83% ±1.11% ±1.46%events/eventtarget-add-remove.js nListener=5 n=1000000 *** 6.41 % ±1.54% ±2.05% ±2.67%events/eventtarget.js listeners=1 n=1000000 *** 259.34 % ±2.83% ±3.81% ±5.05%events/eventtarget.js listeners=10 n=1000000 *** 176.98 % ±1.97% ±2.65% ±3.52%events/eventtarget.js listeners=5 n=1000000 *** 219.14 % ±2.20% ±2.97% ±3.94%
HTTP
The HTTP Servers are one of the most impactful layers of improvement in Node.js. It isn’t a myth that most Node.jsapplications nowadays run an HTTP Server. So, any change can be easily considered a semver-major and increase theefforts for a compatible improvement in performance.
Therefore, the HTTP server utilized is an http.Server
that replies 4 chunks of 256 bytes each containing ‘C’ on eachrequest, as you can see in this example:
http.createServer((req, res) => { const n_chunks = 4; const body = 'C'.repeat(); const len = body.length;res.writeHead(200, {'Content-Type': 'text/plain', 'Content-Length': len.toString()}); for (i = 0, n = (n_chunks - 1); i < n; ++i) res.write(body.slice(i * step, i * step + step)); res.end(body.slice((n_chunks - 1) * step));})// See: https://github.com/nodejs/node/blob/main/benchmark/fixtures/simple-http-server.js
When comparing the performance of Node.js 16 and Node.js 18, there is a noticeable 8% improvement. However, upgradingfrom Node.js 18 to Node.js 20 resulted in a significant improvement of 96.13%.
These benchmark results were collected usingtest-double-http
benchmarker method.Which is, a simple Node.js script to send HTTP GET requests:
function run() { if (http.get) { // HTTP or HTTPS if (options) { http.get(url, options, request); } else { http.get(url, request); } } else { // HTTP/2 const client = http.connect(url); client.on('error', () => {}); request(client.request(), client); }}run();
By switching to more reliable benchmarking tools such as autocannon
or wrk
, we observed a significant drop in thereported improvement — from 96% to 9%.This indicates that the previous benchmarking method had limitations or errors.However, the actual performance of the HTTP server has improved, and we need to carefully evaluate the percentage ofimprovement with the new benchmarking approach to accurately assess the progress made.
Should I expect a 96%/9% performance improvement in my Express/Fastify application?
Absolutely, not. Frameworks may opt not to use the internal HTTP API — that’s one of the reasons Fastify is… fast! Forthis reason, another benchmark suite was considered in this report (3. HTTP Servers).
Misc
According to our tests, the startup.js
script has demonstrated a significant improvement in the Node.js processlifecycle, with a 27% boost observed from Node.js version 18 to version 20. This improvement is even more impressivewhen compared to Node.js version 16, where the startup time was reduced by 34.75%!
As modern applications increasingly rely on serverless systems, reducing startup time has become a crucial factor inimproving overall performance. It’s worth noting that the Node.js team is always working towards optimizing this aspectof the platform, as evidenced by our strategic initiative: https://github.com/nodejs/node/issues/35711.
These improvements in startup time not only benefit serverless applications but also enhance the performance of otherNode.js applications that rely on quick boot-up times. Overall, these updates demonstrate the Node.js team’s commitmentto enhancing the platform’s speed and efficiency for all users.
$ node-benchmark-compare compare-misc-16-18.csv confidence improvement accuracy (*) (**) (***)misc/startup.js count=30 mode='process' script='benchmark/fixtures/require-builtins' *** 12.99 % ±0.14% ±0.19% ±0.25%misc/startup.js count=30 mode='process' script='test/fixtures/semicolon' *** 5.88 % ±0.15% ±0.20% ±0.26%misc/startup.js count=30 mode='worker' script='benchmark/fixtures/require-builtins' *** 5.26 % ±0.14% ±0.19% ±0.25%misc/startup.js count=30 mode='worker' script='test/fixtures/semicolon' *** 3.84 % ±0.15% ±0.21% ±0.27%$ node-benchmark-compare compare-misc-18-20.csv confidence improvement accuracy (*) (**) (***)misc/startup.js count=30 mode='process' script='benchmark/fixtures/require-builtins' *** -4.80 % ±0.13% ±0.18% ±0.23%misc/startup.js count=30 mode='process' script='test/fixtures/semicolon' *** 27.27 % ±0.22% ±0.29% ±0.38%misc/startup.js count=30 mode='worker' script='benchmark/fixtures/require-builtins' *** 7.23 % ±0.21% ±0.28% ±0.37%misc/startup.js count=30 mode='worker' script='test/fixtures/semicolon' *** 31.26 % ±0.33% ±0.44% ±0.58%
This benchmark is pretty straightforward. We measure the time elapsed when creating a new [mode] using the given[script] where [mode] can be:
process
- a new Node.js processworker
- a Node.js worker_thread
And [script] is divided into:
benchmark/fixtures/require-builtins
- a script that requires all the Node.js modulestest/fixtures/semicolon
- an empty script — containing a single;
(semicolon)
This experiment can be easily reproducible with hyperfine
or time
:
$ hyperfine --warmup 3 './node16 ./nodejs-internal-benchmark/semicolon.js'Benchmark 1: ./node16 ./nodejs-internal-benchmark/semicolon.js Time (mean ± σ): 24.7 ms ± 0.3 ms [User: 19.7 ms, System: 5.2 ms] Range (min … max): 24.1 ms … 25.6 ms 121 runs$ hyperfine --warmup 3 './node18 ./nodejs-internal-benchmark/semicolon.js'Benchmark 1: ./node18 ./nodejs-internal-benchmark/semicolon.js Time (mean ± σ): 24.1 ms ± 0.3 ms [User: 18.1 ms, System: 6.3 ms] Range (min … max): 23.6 ms … 25.3 ms 123 runs$ hyperfine --warmup 3 './node20 ./nodejs-internal-benchmark/semicolon.js'Benchmark 1: ./node20 ./nodejs-internal-benchmark/semicolon.js Time (mean ± σ): 18.4 ms ± 0.3 ms [User: 13.0 ms, System: 5.9 ms] Range (min … max): 18.0 ms … 19.7 ms 160 runs
💡 The warmup is necessary to consider the influence of the file system cache
The trace_events
module has also undergone a notable performance boost, with a 7% improvement observed whencomparing Node.js version 16 to version 20. It’s worth noting that this improvement was slightly lower, at 2.39%,when comparing Node.js version 18 to version 20.
Module
require()
(or module.require
) has long been a culprit of slow Node.js startup times. However, recent performanceimprovements suggest that this function has been optimized as well. Between Node.js versions 18 and 20, we observedimprovements of 4.20% when requiring .js
files, 6.58% for .json
files, and 9.50% when readingdirectories - all of which contribute to faster startup times.
Optimizing require()
is crucial because it is a function that’s used heavily in Node.js applications. By reducing thetime it takes for this function to execute, we can significantly speed up the entire startup process and improve theuser experience.
Streams
Streams are an incredibly powerful and widely used feature of Node.js. However, between Node.js versions 16 and 18, someoperations related to streams became slower. This includes creating and destroying Duplex
, Readable
, Transform
,and Writable
streams, as well as the .pipe()
method for Readable → Writable streams.
The graph below illustrates this regression:
However, this pipe
regression was reduced in Node.js 20:
$ node-benchmark-compare compare-streams-18-20.csv confidence improvement accuracy (*) (**) (***)streams/creation.js kind='duplex' n=50000000 *** 12.76 % ±4.30% ±5.73% ±7.47%streams/creation.js kind='readable' n=50000000 *** 3.48 % ±1.16% ±1.55% ±2.05%streams/creation.js kind='transform' n=50000000 ** -7.59 % ±5.27% ±7.02% ±9.16%streams/creation.js kind='writable' n=50000000 *** 4.20 % ±0.87% ±1.16% ±1.53%streams/destroy.js kind='duplex' n=1000000 *** -6.33 % ±1.08% ±1.43% ±1.87%streams/destroy.js kind='readable' n=1000000 *** -1.94 % ±0.70% ±0.93% ±1.21%streams/destroy.js kind='transform' n=1000000 *** -7.44 % ±0.93% ±1.24% ±1.62%streams/destroy.js kind='writable' n=1000000 0.20 % ±1.89% ±2.52% ±3.29%streams/pipe.js n=5000000 *** 87.18 % ±2.58% ±3.46% ±4.56%
And as you may have noticed, some types of streams (Transform
specifically) are regressed in Node.js 20. Therefore,Node.js 16 still has the fastest streams — for this specific benchmark, please do not read this benchmark result as‘Node.js streams in v18 and v20 are so slow!’ This is a specific benchmark that may or may not affect your workload. Forinstance, if you look at a naive comparisonin the nodejs-bench-operations,you will see that the following snippet performs better on Node.js 20 than its predecessors:
suite.add('streams.Writable writing 1e3 * "some data"', function () { const writable = new Writable({ write (chunk, enc, cb) { cb() } }) let i = 0 while(i < 1e3) { writable.write('some data') ++i }})
The fact is, the instantiation and destroy methods play an important role in the Node.js ecosystem. Hence, it’s verylikely to have a negative impact on some libraries. However, this regression isbeing monitored closely in theNode.js Performance WG.
Note that the readable async iterator becomes slightly faster (~6.14%) on Node.js 20.
URL
Since Node.js 18, a new URL parser dependency was added to Node.js — Ada. Thisaddition bumped the Node.js performance when parsing URLs to a new level. Some results could reach up to an improvementof 400%. As a regular user, you may not use it directly. But if you use an HTTP server then it’s very likely to beaffected by this performance improvement.
The URL benchmark suite is pretty large. For this reason, only WHATWG URL benchmark results will be covered.
url.parse()
and url.resolve()
are both deprecated and legacy APIs. Even though its usage is considered a risk forany Node.js application, developers still use it. Quoting Node.js documentation:
url.parse()
uses a lenient, non-standard algorithm for parsing URL strings. It is prone to security issues suchashost name spoofingand incorrect handling of usernames and passwords. Donot use with untrusted input. CVEs are not issued forurl.parse()
vulnerabilities. UsetheWHATWG URLAPI instead.
If you are curious about the performance changes of url.parse
and url.resolve
, check out theState of Node.js Performance 2023 repository.
That said, it’s really interesting to see the results of the new whatwg-url-parse:
Below is a list of URLs used for benchmarking, which were selected based on the benchmark configuration
const urls = { long: 'http://nodejs.org:89/docs/latest/api/foo/bar/qua/13949281/0f28b/' + '/5d49/b3020/url.html#test?payload1=true&payload2=false&test=1' + '&benchmark=3&foo=38.38.011.293&bar=1234834910480&test=19299&3992&' + 'key=f5c65e1e98fe07e648249ad41e1cfdb0', short: 'https://nodejs.org/en/blog/', idn: 'http://你好你好.在线', auth: 'https://user:[emailprotected]/path?search=1', file: 'file:///foo/bar/test/node.js', ws: 'ws://localhost:9229/f46db715-70df-43ad-a359-7f9949f39868', javascript: 'javascript:alert("node is awesome");', percent: 'https://%E4%BD%A0/foo', dot: 'https://example.org/./a/../b/./c',}
With the recent upgrade of Ada 2.0 in Node.js 20, it’s fair to say there’s also a significant improvement when comparingNode.js 18 to Node.js 20:
And the benchmark file is pretty simple:
function useWHATWGWithoutBase(data) { const len = data.length; let result = new URL(data[0]); // Avoid dead code elimination bench.start(); for (let i = 0; i < len; ++i) { result = new URL(data[i]); } bench.end(len); return result;}function useWHATWGWithBase(data) { const len = data.length; let result = new URL(data[0][0], data[0][1]); // Avoid dead code elimination bench.start(); for (let i = 0; i < len; ++i) { const item = data[i]; result = new URL(item[0], item[1]); } bench.end(len); return result;}
The only difference is the second parameter that is used as a base when creating/parsing the URL. It’s also worthmentioning that when a base is passed (withBase=’true’), it tends to perform faster than the regular usage(new URL(data)
). See all the results expanded inthe main repository.
Buffers
In Node.js, buffers are used to handle binary data. Buffers are a built-in data structure that can be used to store rawbinary data in memory, which can be useful when working with network protocols, file system operations, or otherlow-level operations. Overall, buffers are an important part of Node.js and are used extensively throughout the platformfor handling binary data.
For those of you who make use directly or indirectly of Node.js buffers, I have good news (mainly for Node.js 20 earlyadopters).
Besides improving the performance of Buffer.from()
Node.js 20 fixed two main regressions from Node.js 18:
Buffer.concat()
Node.js version 20 has shown significant improvements compared to version 18, and these improvements remain apparenteven when compared to version 16:
Buffer.toJSON()
From Node.js 16 to Node.js 18, a drop of 88% in the performance of Buffer.toJSON
was observed:
$ node-benchmark-compare compare-buffers-16-18.csv confidence improvement accuracy (*) (**) (***)buffers/buffer-tojson.js len=256 n=10000 *** -81.12 % ±1.25% ±1.69% ±2.24%buffers/buffer-tojson.js len=4096 n=10000 *** -88.39 % ±0.69% ±0.93% ±1.23%
However, this regression was fixed and improved in Node.js 20 by orders of magnitude!
$ node-benchmark-compare compare-buffers-18-20.csv confidence improvement accuracy (*) (**) (***)buffers/buffer-tojson.js len=256 n=10000 *** 482.81 % ±7.02% ±9.42% ±12.42%buffers/buffer-tojson.js len=4096 n=10000 *** 763.34 % ±5.22% ±7.04% ±9.34%
Therefore, it’s correct to state that Node.js 20 is the fastest version of Node.js in dealing with buffers.
See the full comparison between Node.js 20 and Node.js 18 below:
$ node-benchmark-compare compare-buffers-18-20.csv confidence improvement accuracy (*) (**) (***)buffers/buffer-base64-decode.js size=8388608 n=32 *** 1.66 % ±0.10% ±0.14% ±0.18%buffers/buffer-base64-encode.js n=32 len=67108864 *** -0.44 % ±0.17% ±0.23% ±0.30%buffers/buffer-compare.js n=1000000 size=16 *** -3.14 % ±0.82% ±1.09% ±1.41%buffers/buffer-compare.js n=1000000 size=16386 *** -15.56 % ±5.97% ±7.95% ±10.35%buffers/buffer-compare.js n=1000000 size=4096 -2.63 % ±3.09% ±4.11% ±5.35%buffers/buffer-compare.js n=1000000 size=512 *** -6.15 % ±1.28% ±1.71% ±2.24%buffers/buffer-concat.js n=800000 withTotalLength=0 pieceSize=1 pieces=16 *** 300.67 % ±0.71% ±0.95% ±1.24%buffers/buffer-concat.js n=800000 withTotalLength=0 pieceSize=1 pieces=4 *** 212.56 % ±4.81% ±6.47% ±8.58%buffers/buffer-concat.js n=800000 withTotalLength=0 pieceSize=16 pieces=16 *** 287.63 % ±2.47% ±3.32% ±4.40%buffers/buffer-concat.js n=800000 withTotalLength=0 pieceSize=16 pieces=4 *** 216.54 % ±1.24% ±1.66% ±2.17%buffers/buffer-concat.js n=800000 withTotalLength=0 pieceSize=256 pieces=16 *** 38.44 % ±1.04% ±1.38% ±1.80%buffers/buffer-concat.js n=800000 withTotalLength=0 pieceSize=256 pieces=4 *** 91.52 % ±3.26% ±4.38% ±5.80%buffers/buffer-concat.js n=800000 withTotalLength=1 pieceSize=1 pieces=16 *** 192.63 % ±0.56% ±0.74% ±0.97%buffers/buffer-concat.js n=800000 withTotalLength=1 pieceSize=1 pieces=4 *** 157.80 % ±1.52% ±2.02% ±2.64%buffers/buffer-concat.js n=800000 withTotalLength=1 pieceSize=16 pieces=16 *** 188.71 % ±2.33% ±3.12% ±4.10%buffers/buffer-concat.js n=800000 withTotalLength=1 pieceSize=16 pieces=4 *** 151.18 % ±1.13% ±1.50% ±1.96%buffers/buffer-concat.js n=800000 withTotalLength=1 pieceSize=256 pieces=16 *** 20.83 % ±1.29% ±1.72% ±2.25%buffers/buffer-concat.js n=800000 withTotalLength=1 pieceSize=256 pieces=4 *** 59.13 % ±3.18% ±4.28% ±5.65%buffers/buffer-from.js n=800000 len=100 source='array' *** 3.91 % ±0.50% ±0.66% ±0.87%buffers/buffer-from.js n=800000 len=100 source='arraybuffer-middle' *** 11.94 % ±0.65% ±0.86% ±1.13%buffers/buffer-from.js n=800000 len=100 source='arraybuffer' *** 12.49 % ±0.77% ±1.03% ±1.36%buffers/buffer-from.js n=800000 len=100 source='buffer' *** 7.46 % ±1.21% ±1.62% ±2.12%buffers/buffer-from.js n=800000 len=100 source='object' *** 12.70 % ±0.84% ±1.12% ±1.47%buffers/buffer-from.js n=800000 len=100 source='string-base64' *** 2.91 % ±1.40% ±1.88% ±2.46%buffers/buffer-from.js n=800000 len=100 source='string-utf8' *** 12.97 % ±0.77% ±1.02% ±1.33%buffers/buffer-from.js n=800000 len=100 source='string' *** 16.61 % ±0.71% ±0.95% ±1.25%buffers/buffer-from.js n=800000 len=100 source='uint16array' *** 5.64 % ±0.84% ±1.13% ±1.48%buffers/buffer-from.js n=800000 len=100 source='uint8array' *** 6.75 % ±0.95% ±1.28% ±1.68%buffers/buffer-from.js n=800000 len=2048 source='array' 0.03 % ±0.33% ±0.43% ±0.56%buffers/buffer-from.js n=800000 len=2048 source='arraybuffer-middle' *** 11.73 % ±0.55% ±0.74% ±0.96%buffers/buffer-from.js n=800000 len=2048 source='arraybuffer' *** 12.85 % ±0.55% ±0.73% ±0.96%buffers/buffer-from.js n=800000 len=2048 source='buffer' *** 7.66 % ±1.28% ±1.70% ±2.21%buffers/buffer-from.js n=800000 len=2048 source='object' *** 11.96 % ±0.90% ±1.20% ±1.57%buffers/buffer-from.js n=800000 len=2048 source='string-base64' *** 4.10 % ±0.46% ±0.61% ±0.79%buffers/buffer-from.js n=800000 len=2048 source='string-utf8' *** -1.30 % ±0.71% ±0.96% ±1.27%buffers/buffer-from.js n=800000 len=2048 source='string' *** -2.23 % ±0.93% ±1.25% ±1.64%buffers/buffer-from.js n=800000 len=2048 source='uint16array' *** 6.89 % ±1.44% ±1.91% ±2.49%buffers/buffer-from.js n=800000 len=2048 source='uint8array' *** 7.74 % ±1.36% ±1.81% ±2.37%buffers/buffer-tojson.js len=0 n=10000 *** -11.63 % ±2.34% ±3.11% ±4.06%buffers/buffer-tojson.js len=256 n=10000 *** 482.81 % ±7.02% ±9.42% ±12.42%buffers/buffer-tojson.js len=4096 n=10000 *** 763.34 % ±5.22% ±7.04% ±9.34%
Text Encoding and Decoding
TextDecoder and TextEncoder are two JavaScript classes that are part of the Web APIs specification and are available inmodern web browsers and Node.js. Together, TextDecoder and TextEncoder provide a simple and efficient way to work withtext data in JavaScript, allowing developers to perform various operations involving strings and character encodings.
Decoding and Encoding becomes considerably faster than in Node.js 18. With the addition ofsimdutf for UTF-8 parsing the observed benchmark, results improved by 364%(an extremely impressive leap) when decoding in comparison to Node.js 16.
Those improvements got even better on Node.js 20, with a performance improvement of 25% in comparison to Node.js 18.See the full results in thestate-of-nodejs-performance-2023 repository.
Performance improvements were also observed when comparing encoding methods on Node.js 18. From Node.js 16 to Node.js18, the TextEncoder.encodeInto
reached 93.67% of improvement in the current observation (using ascii
with astring length of 256):
Node.js Bench Operations
The benchmarking operations in Node.js have always piqued my curiosity. As someone who enjoys exploring the intricaciesof Node.js and its underlying technology, I find it fascinating to delve into the details of these operations,particularly those related to the V8 engine. In fact, I often like to share my findings with others through talks andworkshops delivered by NearForm, a company I’m affiliated with. If you’re interested, youcan find more information about my presentations on this topic by clicking this link.
In addition, these benchmarks will use the ops/sec
metric, which basically means the number of operations that wereperformed in one second. It’s important to emphasize that this can only mean a very small fraction of your computingtime. If you have read my previous article(Preparing and Evaluating Benchmarks)you should remember the ‘Evaluating Results’ section, where I approach the problem with ops/sec
in real-worldapplications — if not, you should consider returning to it.
Parsing Integers
Parsing strings to numbers can be accomplished using either +
or parseInt(x, 10)
. Previous benchmark resultsshowed that using +
was faster than parseInt(x, 10)
in earlier versions of Node.js, as illustrated in the tablebelow:
name | ops/sec | samples |
---|---|---|
Using parseInt(x, 10) - small number (2 len) | 283,768,532 | 91 |
Using parseInt(x, 10) - big number (10 len) | 21,307,115 | 100 |
Using + - small number (2 len) | 849,906,952 | 100 |
Using + - big number (10 len) | 849,173,336 | 97 |
However, with the release of Node.js 20 and the new V8 version (11.4), both operations have become equivalent in termsof performance, as shown in the updated benchmark results below:
name | ops/sec | samples |
---|---|---|
Using parseInt(x, 10) - small number (2 len) | 856,413,575 | 98 |
Using parseInt(x, 10) - big number (10 len) | 856,754,259 | 96 |
Using + - small number (2 len) | 857,364,191 | 98 |
Using + - big number (10 len) | 857,511,971 | 96 |
Super vs This
One of the interesting benchmarks that have changed with the addition of Node.js 20 is the usage of this
or super
inclasses, as you can see in the example underneath:
class Base { foo () { return 10 * 1e2 }}class SuperClass extends Base { bar () { const tmp = 20 * 23 return super.foo() + tmp }}class ThisClass extends Base { bar () { const tmp = 20 * 23 return this.foo() + tmp }}
The comparison between super
and this
in Node.js 18 was producing the following operations per second (ops/sec):
name | ops/sec | samples |
---|---|---|
Using super | 159,426,608 | 96 |
Using this | 160,092,440 | 91 |
There isn’t a significant difference between both approaches and on Node.js 20. This statement holds with a slightdifference:
name | ops/sec | samples |
---|---|---|
Using super | 850,760,436 | 97 |
Using this | 853,619,840 | 99 |
Based on the benchmark results, it appears that there has been a significant increase in performance when using this
on Node.js 20 compared to Node.js 18. This increase is quite remarkable, with this
achieving an impressive853,619,840 ops/sec on Node.js 20 compared to only 160,092,440 ops/sec on Node.js 18, which is, 433% better!Apparently, it has the same property access method as a regular object: obj.property1
. Also, note that both operationswere tested in the same dedicated environment. Therefore, it’s unlikely to have occurred by chance.
Property Access
There are various ways to add properties to objects in JavaScript, each with its own purpose and sometimes ambiguous innature. As a developer, you may wonder about the efficiency of property access in each of these methods.
The good news is that the nodejs-bench-operations repository includes a comparison of these methods, which sheds lighton their performance characteristics. In fact, this benchmarking data reveals that the property access in Node.js 20 hasseen significant improvements, particularly when using objects with writable: true
andenumerable/configurable: false
properties.
const myObj = {};Object.defineProperty(myObj, 'test', { writable: true, value: 'Hello', enumerable: false, configurable: false,});myObj.test // How fast is the property access?
On Node.js 18 the property access (myObj.test) was producing 166,422,265 ops/sec. However, under thesame circ*mstances, Node.js 20 is producing 857,316,403 ops/sec! This and other particularities around property accesscan be found in the following benchmark results:
- Property getter accessv18 /v20
- Property setter accessv18 /v20
- Property access after shape transitionv18 /v20
Array.prototype.at
Array.prototype.at(-1)
is a method that was introduced in the ECMAScript 2021 specification. It allows you to accessthe last element of an array without knowing its length or using negative indices, which can be a useful feature incertain use cases. In this way, the at()
method provides a more concise and readable way to access the last element ofan array, compared to traditional methods like array[array.length - 1]
.
On Node.js 18 this access was considerably slower in comparison to Array[length-1]
:
name | ops/sec | samples |
---|---|---|
Length = 100 - Array.at | 26,652,680 | 99 |
Length = 10_000 - Array.at | 26,317,564 | 97 |
Length = 1_000_000 - Array.at | 27,187,821 | 98 |
Length = 100 - Array[length - 1] | 848,118,011 | 98 |
Length = 10_000 - Array[length - 1] | 847,958,319 | 100 |
Length = 1_000_000 - Array[length - 1] | 847,796,498 | 101 |
Since Node.js 19, Array.prototype.at is equivalent to theold-fashioned Array[length-1] as the table below suggests:
name | ops/sec | samples |
---|---|---|
Length = 100 - Array.at | 852,980,778 | 99 |
Length = 10_000 - Array.at | 854,299,272 | 99 |
Length = 1_000_000 - Array.at | 853,374,694 | 98 |
Length = 100 - Array[length - 1] | 854,589,197 | 95 |
Length = 10_000 - Array[length - 1] | 856,122,244 | 95 |
Length = 1_000_000 - Array[length - 1] | 856,557,974 | 99 |
String.prototype.includes
Most people know that RegExp is very often the source of many bottlenecks in any kind of application. For instance,you might want to check if a certain variable contains application/json
.And while you can do it in several manners,most of the time you will end up using either:
/application\/json/.test(text)
- RegEx
or
text.includes('application/json')
- String.prototype.includes
What some of you may not know is that String.prototype.includes
is pretty much as slow as RegExp on Node.js 16.
name | ops/sec | samples |
---|---|---|
Using includes | 16,056,204 | 97 |
Using indexof | 850,710,330 | 100 |
Using RegExp.test | 15,227,370 | 98 |
Using RegExp.text with cached regex pattern | 15,808,350 | 97 |
Using new RegExp.test | 4,945,475 | 98 |
Using new RegExp.test with cached regex pattern | 5,944,679 | 100 |
However, since Node.js 18 this behavior was fixed.
name | ops/sec | samples |
---|---|---|
Using includes | 856,127,951 | 101 |
Using indexof | 856,709,023 | 98 |
Using RegExp.test | 16,623,756 | 98 |
Using RegExp.text with cached regex pattern | 16,952,701 | 99 |
Using new RegExp.test | 4,704,351 | 95 |
Using new RegExp.test with cached regex pattern | 5,660,755 | 95 |
Crypto.verify
In Node.js, the crypto module provides a set of cryptographic functionalities that can be used for various purposes,such as creating and verifying digital signatures, encrypting and decrypting data, and generating secure random numbers.One of the methods available in this module is crypto.verify()
, which is used to verify a digital signature generatedby the crypto.sign()
method.
Node.js 14 (End-of-Life) uses OpenSSL 1.x. On Node.js 16 we’ve had the addition of theQUIC protocol, but still using OpenSSL version 1. However, in Node.js 18 we’veupdated OpenSSL to version 3.x (over QUIC), and aregression was found after Node.js 18 that reduced from 30k ops/secto 6~7k ops/sec. As I’ve mentioned in the tweet, it’s verylikely to be caused by the new OpenSSL version. Again, our team is looking into it and if you have any insight on this,feel free to comment on the issue: https://github.com/nodejs/performance/issues/72.
Node.js performance initiatives
The Node.js team has always been careful to ensure that its APIs and core functionalities are optimized for speed andresource usage.
In order to further enhance the performance of Node.js, the team has recentlyintroduced a new strategic initiative called ‘Performance’,which is chaired by Yagiz Nizipli. This initiative is aimed at identifying and addressingperformance bottlenecks in the Node.js runtime and core modules, as well as improving the overall performance andscalability of the platform.
In addition to the Performance initiative, there are several other initiatives currently underway that are focused onoptimizing different aspects of Node.js. One of these initiatives is the‘Startup Snapshot’ initiative, which is chaired byJoyee. This initiative is aimed at reducing the startup time of Node.jsapplications, which is a critical factor in improving the overall performance and user experience of web applications.
Therefore, if you are interested in this subject, consider joining the meetings every other week, and feel free to senda message in the #nodejs-core-performance
channel on theOpenJS Foundation Slack.
Things to keep an eye on
Besides the strategic initiatives, there are some pull requests that are very likely to have a great impact on theNode.js performance — at the moment I’m writing the below post (it isn’t merged yet):
- Node.js Errors - https://github.com/nodejs/node/pull/46648
Errors are very expensive to create in Node.js. It’s very often a source of bottlenecks in Node.js applications. As anexample,I conducted research on the implementation of fetch in Node.js(undici) and discovered one of the villains in the Node.js WebStreams implementation is error creation. Hence, byoptimizing error objects in Node.js, we can improve the overall efficiency of the platform and reduce the risk ofbottlenecks.
- Pointer compression builds - https://github.com/nodejs/build/issues/3204
Pointer compression is a technique used in computer programming to reduce the memory usage of programs that make use ofmany pointers. While it doesn’t improve performance directly, it can indirectly improve performance by reducing cachemisses and page faults. This certainlycan reduce some infra costs, as described in the issue thread.
- Increase default
--max-semi-space-size
- https://github.com/nodejs/node/pull/47277
An issue was created in March 2022 suggesting increasing the V8max_semi_space_size
with the objective to reduce the Garbage Collection (Scavenge specifically) runs and increasingthe overall throughput in the web tooling benchmark. We’re still evaluating its impact and it may or may not arrive inNode.js 21.
- bump
highWaterMark
value on Node.js Readable/Writable streams - https://github.com/nodejs/node/pull/46608
This PR increases the default value for highWaterMark
value in Node.js streams. It’s expected to perceive aperformance improvement in the Node.js stream usage with default options. This PR however, is a semver-major
changeand should arrive on Node.js 21. For a detailed benchmark result, wait for: ‘State of Node.js Performance 2023 - P2’ atthe end of the year.
Conclusion
Despite some regressions in the Node.js streams and crypto module, Node.js 20 boasts significant improvements inperformance compared to previous versions. Notable enhancements have been observed in JavaScript operations such asproperty access, URL parsing, buffers/text encoding and decoding, startup/process lifecycle time, and EventTarget, amongothers.
The Node.js performance team(nodejs/performance) hasexpanded its scope, leading to greater contributions in optimizing performance with each new version. This trendindicates that Node.js will continue to become faster over time.
It’s worth mentioning that the benchmark tests focus on specific operations, which may or may not directly impact yourspecific use case. Therefore, I strongly recommend reviewing all the benchmark results in thestate-of-nodejs-performance repository and ensuringthat these operations align with your business requirements.
Acknowledgments
I would like to express my sincere gratitude to all the reviewers who took the time to provide valuable feedback on myblog post. Thank you for your time, expertise, and constructive comments.
It’s also worth noting that I continue to contribute to open-source projects in my free time out of love for the community.If my work has positively impacted you and you’d like to express your appreciation, consider sponsoring me onGitHub 💚.