Node.js vs Go (Golang): Performance (Latency - Throughput - Saturation - Availability)

[@AntonPutra]

► What should I test next? ► AWS is expensive - Infra Support Fund: buymeacoffee.com/antonputra ► Benchmarks: kzbin.info/aero/PLiMWaCMwGJXmcDLvMQeORJ-j_jayKaLVn&si=p-UOaVM_6_SFx52H

[@Yukiixs]

UDP echo server!

[@alexreustle]

Would you consider OLAP database benchmarks? Compare managed RDS postgres vs Redshift vs self managed greenplum cluster, vs clickhouse, duckdb, monetdb etc?

[@朝秦-f2h]

node vs bun vs deno

[@adriandrozman3404]

fastify vs elysia vs deno

[@efrainsoteloferry4075]

Go(stdlib) vs FastApi(Python)

[@Blue-jk4fw]

Can we see Bun vs Golang?

[@Blue-jk4fw]

I recommend Hono as the backend

[@Abioticodin2387]

@@Blue-jk4fw or Elysia since it's a bun-specific library

[@sherkan_n]

@@Blue-jk4fwor elysia

[@bhuvanbm9104]

Hono with bun is great and soo simple

[@AlexanderMoon]

hyper-express (uWebSockets) will outperform bun/deno with express.

[@Gorops]

Huh, I wonder how the newest JS runtimes, Deno 2 and Bun, compare to the classic Node.

[@JohnDoe-ji1zv]

What’s the point to compare them if nobody using them in production anyway

[@lucsoft]

@@JohnDoe-ji1zvmany companies use deno

[@Serizon_]

Please make a video on first comparing bun js and deno 2 speed then bun and golang I REALLY LOVE YOUR VIDEOS

[@artemcodes]

Amazing approach with the metrics. It’s was very nice to see the difference on real environment. Keep it going!

[@AntonPutra]

thank you!

[@conman7644]

yeah, these newer runtimes would be really interesting to see. like bun and deno 2.0

[@AntonPutra]

next deno vs. node.js vs bun 😊

[@conman7644]

​@@AntonPutra sounds awesome

[@midewestmond9442]

​@@AntonPutralet go

[@AGAU1022]

​@@AntonPutraIf you use Elysia for Bun, then I would recommend using Fresh for Deno and Fastify or Hono for Node to keep it fair.

[@ehfoss]

Babe, wake up. Anton posted more benchies!

[@Mr.BinarySniper]

would be awesome video.

[@gustavomelo6560]

Amazing benchmarks as always you do! Please, make a video teaching us how to setup kubernets, grafana and prometheus to start play with some codes. Maybe a minikube, sidecards, deploying apps, metrics, etc. I would love to start playing with all of this tecniques

[@fuzzylogicq]

I think something like a leader board for all the benchmarks would be interesting, (like how car drags have some board or laptime boards etc) would be interesting to come back later and see what tops and what tradeoffs to make between a certain language or framework etc.

[@AntonPutra]

thanks! i was thinking about it. i want to eventually compare the fastest frameworks that each language has, and maybe also compare interpreted and compiled ones, then have a final battle 😊

[@jungervin8765]

The difference is crazy I have never thought the difference can be this much, thank you. I made a good decision to use go :D

[@galower405]

Express is still one of the slowest frameworks for node, you may want to give a try to something like Hono.js. Also perhaps using something like the deno or the bun runtime which provide massive performance difference.

[@AntonPutra]

yes, it's just a beginning.. next: deno vs. node.js vs bun

[@jhirn2957]

These are fun to watch. I’d love to see one with Rails/Puma vs. whatever. I know Ruby isn’t going to win any drag races but it’s still the most Rapid development experience by a mile. I’d be happy to provide assistance with any configuration if you need.

[@exapsy]

i love you man, thanks for these.

[@AntonPutra]

thanks you! :)

I would love to see the comparison of Java frameworks: helidon vs spring boot vs micronaut vs quarkus - with and without the support of native images. Many thanks for the great job you already have done!

[@baluditor]

Thank you! Exactly what I wanted to see :)

[@anonymous2596]

You are doing gods work

[@sumantpatel7262]

Great job Anton, we are getting very insightful info on frameworks and runtime, but most of us in community are familiar with JS, can you please provide the comparison of different JS runtime maybe express but in bun runtime.

[@thisistotest]

Sir, C++ vs Rust vs Go. Especially C++ vs Rust.

[@bloodpriest1302]

Nothing beats rust. Every other thing competes fr the 2nd place

[@thisistotest]

@bloodpriest1302 zig is faster than rust.

[@bloodpriest1302]

@@thisistotest no its not. particular zig implementations of something could be faster under some scenarios involving dealing with default workings of both, but once you place in the optimizations & everything, rust is better

[@ManiBalajiC]

man throwing those instances like its cheap ...

[@Serizon_]

I am noob , could you please tell me how much it would cost? I don't think it would cost that much as compared to the reputation he has garnered in the industry

[@asdasfdfgewqgrgyjh]

​@@Serizon_ Youre right its not that much as he was only running the test for two hours. You can use AWS on demand pricing chart to figure it our. He ran the tests for two hours and used the following EC2 nodes. 3 m7g.4xlarge: $0.6528 per node per hour = $3.9168 1 m7g.2xlarge: $0.3264 per node per hour = $0.6528 2 m7g.xlage: $0.0816 per node per hour = $0.3264 For a total of $4.896 for the full test. Now he was running the nodes for longer then the test period but that should be a good estimate of how much this test cost him. Oh and you'd need to add however much the EKS cost to run in AWS that

[@asdasfdfgewqgrgyjh]

Look like I only counted the time for the second test. So you'll need to add another hour for the first test

[@JonathanFraser-i7h]

@@asdasfdfgewqgrgyjh EKS is only like $100 bucks a month for the hosted service. It's basically zero compared to compute unless you are running really small nodes.

[@MS-oy3ry]

@@asdasfdfgewqgrgyjh and why exactly would you use on demand instead of spot?

Wonderful and simple content , thank you !!

[@AntonPutra]

thank you!

[@BrunoFerro-sy9en]

Great work man! But I think that we need Java vs C# and Node Vs Python comparison too.

[@Mr.BinarySniper]

Sir, you've promised your next video will be Node vs Bun vs Deno we are waiting for that legendary js compression.

[@AntonPutra]

it's literally the next video, which i hope to release in a couple of days: "deno vs. node.js vs bun" 😊

[@simonparis84]

You should properly configure the connection backlog size in node. If you don't, the server is spending most of its time on requests that are already very very close to the timeout limit. And, it looks like your client latency chart doesnt count timed out requests which gives a very wrong picture.

[@skuwamy]

🤣🤣🤣🤣🤣

[@yuu.relief]

Its so funny when i read notification "Anton Putra publish a new video". Your channel name sound like Southeast asia people. btw nice comparation, i never expect node.js vcu stuck at around 60%

[@musashigarami]

Not surprised at all, but very interesting to see! Would be good to have a Node Vs Bun Vs Deno 2.0

[@Mr.BinarySniper]

he promised next video will be Node vs Bun vs Deno 2.0

[@AntonPutra]

yes, it will be next in 2 or 3 days; i just decided to start with the slowest one 😊

[@musashigarami]

@@AntonPutra Awesome! This is great content

[@Leonhart_93]

Better results than I thought for NodeJS, especially considering the CPU usage was lower for a lower amount of requests.

[@ThomazMartinez]

pls do Go vs Bun

[@codedusting]

Can you create a tutorial on how to setup such system for comparison? So that we can do it ourselves for any 2 languages.

[@AntonPutra]

yes will it in near future

[@grandlagging0zero175]

Thanks for this video! Maybe. you create a table with all test results(all test

[@AntonPutra]

perhaps in the future, for now, i'll just keep improving the tests. eventually, when i standardize them, i will

[@vikaspoddar001]

Anton brother, you are great

[@AntonPutra]

haha, thanks!

[@xNajdaa]

Would love to see a test of a language on the BEAM that is designed for optimal concurrancy like Elixir or the newer Gleam!

[@errores-me6yu]

дождался, спасибо!

[@AntonPutra]

haha, more js to come 😊

[@cesarmartinez2207]

Hey bro, can you do an express-js(deno) vs express-js(bun) perfomance benchmark, the gain here is to see real life tests to confirm which js runtime is winning, if need help, would be glad to help and thanks again for this amazing work

[@duongphuhiep]

thanks for the hard work. Your videos are very helpful. It will be better if you keep your test scenarios and the infrastructure consistent so that the Go numbers won't changed between your differents videos, then we can "join" all of your videos together and make a table out of them to compare everything together. I mean that it is better to have a "fixed base" though out all of your videos. To simplify things, this "fixed base" would only focus on the max RPS, and the Latency (at 1K rps) for example. The number (max rps, latency) is fixed for Go/stdlib, Viewer should only focus on the results of the "challenger" (Java, Node, C#...) By the way, I'd like to request comparison for the most popular backends technos: Java Quarckus, C#, Node/Deno, PHP vs Go stdlib as the base. I bet that many people would like to see the fight between them :D

[@AntonPutra]

i get your point, but with each new video, i try to improve something, and once in a while, i have to change the infra setup as well. at some point, i'll probably freeze infra changes and focus on the applications. also, thanks for the benchmark suggestions.

[@boot-strapper]

Node sucks, but I highly recommend trying cluster mode. If you want peak js performance try bun with cluster/reuse port turned on. With that setup I found similar performance to golang in my testing

[@badscrew4023]

you need more total CPUs though?

[@boot-strapper]

@@badscrew4023 go already does this, utilizing cluster makes it apples to apples

[@D9ID9I]

no it doesn't sucks.

[@boot-strapper]

@@badscrew4023 nope

[@WhiteRickRoss]

Stop asking why not this framework... He uses the GO standard library and express which is kind of standard for nodejs. Go also has very performant frameworks, but here the situation is to test the standard libraries

[@skuwamy]

Saw it coming, javascript kiddies are delusional.

[@тимур_атмосферный]

Express is slow and was never a Node.js standard

[@Александр-ф9в4ю]

Node.js itself is a joke.

[@skuwamy]

@@тимур_атмосферный lol there is no standard in the JS world.

[@eugenekalashnikov9331]

How you can say that express is a standard?

[@andresaguilar3055]

Thank you!. This is the benchmark Ive been waiting! I always wondered why js kept degrading so fast compared to go even in low stress situationts. It would be really cool to see the impact of the number of database connections allowed in the performance.

[@shadowfaxenator]

You should use gomaxproc equal to the limit of k8s for your pod, otherwise go uses cores equal to the maximum number of cores of host machine, and k8s throttles them. It can lead to a huge performance penalty. The second thing you have to use scale nodejs pods to the number == of golang process cores.

[@ooijaz6063]

Then you will compare 4 nodejs applications vs 1 go application. I think better comparision would be to test performance on machine with only 1 core.

[@Kavantix]

Exactly this, node only uses 1 core + a little for runtime. And Gomaxproc would indeed more reflect a production, although in this case it doesnt really matter 😂

[@lucaerrani3734]

I think (and pls correct me if I'm wrong) that the instances you're using (m7g.4xlarge) have 16vCPUs but since NodeJS is single threaded it cannot leverage more than one. This could be an explanation about why the RPS metrics maxed at 9k/s but the CPU was heavily underutilized (40%ish free)

[@TheRanguna]

So he's only running a single instance per cpu? I thought k8s would scale the pod instances inside the same machine.

[@lucaerrani3734]

@@TheRanguna No he's running a single nodejs instance per pod and one pod per language (one for nodejs/express, one for go) K8s automatically scales the number of pods if configured, but this test setup doesn't seem to have horizontal scaling enabled (which is correct since that would increase dramatically the changing variables in the test) Therefore since go is multithreaded, it is able to use all the vCPUs of the pod it's running on, while nodejs isn't (but that's just my opinion, I quickly skimmed the conf files in the repo and I didn't find anything that would seem to counter it)

[@TheRanguna]

@@lucaerrani3734 took a look at the spec and it seems like each node process is getting 2 vCPU, meaning 1 vCPU is wasted for every node process. Yep, that could explain why node doesn't go further than around 60% cpu usage.

[@uwontlikeit]

💡more ideas for benchmark testing: NATS vs Kafka, Kafka vs Pulsar, NATS microservice communication vs Rest. Great videos, than you for doing it!

[@AntonPutra]

thanks! noted! i'll do on some infra components in the near future

[@commonfolk663]

But express is slowest nodejs we server, you would use fastify or even deno 2.0 for speed

[@skuwamy]

whatever js web framework you throw at go. you will never beat or come near as fast as go. LOL

[@commonfolk663]

​@@skuwamyit doesn't matter, it just makes little sense to compare worst server implementation for nodejs. It's not why you would use nodejs for backend anyway, especially with apps as simple as ud do with golang

[@xomiachuna]

​@@skuwamycan you elaborate on the simplicity of apps? From my understanding go is quite robust and in fact kubernetes is written in it (looks complicated enough for me). I don't have much experience with node or go backends, just curious what you mean by that

[@skuwamy]

@@xomiachuna what the hell you are talking about? 🤣🤣

[@TweakMDS]

I'm loving these language-specific benchmarks, but I would also be super interested in queueing technology comparison. For example: pub/sub using mqtt (Mosquitto for example), Kafka, NATS and maybe some AWS native queueing. Would you be interested in this?

[@AntonPutra]

yes, actually it was my initial idea to test open source databases, messaging systems, etc., but then i got sucked in by the requests for testing different frameworks 😊 i'll get there, i promise

[@user-abc855]

Thank you for the video! Express.js is veeeery slow, what about usage of Fastify with fastify-uws instead?

[@Payamafshari-r7k]

do .NET 8 or 9 with optimization with golang stdlib

[@nixoncode]

.net8 as it's lts

[@gahshunker]

Yeah, do the .NET aot compilation, to be on par with golang

[@alehkhantsevich113]

@@nixoncode LTS doesn't mean anything. I am working in Enterprise and our project will be migrated to .NET 9 in November. So, I would like to see .NET 9 RC NativeAOT vs Go

[@nixoncode]

@@alehkhantsevich113 perhaps you're not in control, jumping on to 9 right now just seems stupid as you're not even compensated for it

[@0xd3c0d3d]

@@nixoncode It has already been announced that NET 9 has huge performance gains compared to the previous version. And this is actually happening, there is a lot of work on internal APIs, especially in Linq.

[@steveoc64]

Would be interesting to see a comparison of the same load running on AWS vs Azure vs GCP vs Some Generic VPS Do you get what you pay for with the big hosting companies?

[@lastink444]

Would be nice to see some alternatives to express that have better performance. Fastify maybe

[@ValerioBarbera]

Fantastic!

[@AntonPutra]

thank you!

[@GolderiQ]

Redo the same with Deno, Bun, Node.js, Go, and Rust please! Please!

[@stickyblicky11]

Theres a video on this

[@GolderiQ]

@@stickyblicky11 I was so happy reading your comment, then I search the name of the channel plus Bun, Deno, Elysia, and Hono and nothing pop up

[@MrEW1985]

As expected nodejs is slaughtered

[@ADP4031]

Yeah, NodeJS was not made to be fast though, results were previsible and fair

[@steinarst9084]

@@ADP4031 a language created over a weekend competing with a massive collaborative superpowered hivemind competing with eachother. of course javascript always lose. but using express for benchmark is just downright cruel.

[@ireydiak]

@@steinarst9084 Just like using the std in Go for the server which is not exactly optimized for this kind of test. Using something like Echo would be more relevant. Also doing something a little bit realistic and involved than a simple post request that generates a uuid.

[@Nurgul626]

Hey, great video! (previous comment was deleted, trying one more time without any links) At a glance, numbers don't add up for node. Yes, node is slower, but not so much, so I checked source code for your tests: 0. Express is really not the fastest guy in town, you can try using fastify for example, or a built-in HTTP module to remove unnecessary overhead 1. I might be wrong, but it looks like go example uses prepared statements for inserts, while node is running the full SQL insert for each request 2. You are using async functions for express, while express doesn't support async middleware (I don't think it affects performance too much, but can be the case for availability) 3. You are giving 1 CPU core to nodejs in kubernetes configuration. Can't tell for Go, but Node uses concurrent Garbage Collection, and giving node process 1 CPU really hits GC times and as a result, the performance 4. Also noticed that in Node example you are modifying payload directly on the request object and pass it into functions, which can give you memory leaks (you don't seem to get one, but it definitely can hurt the performance)

[@AntonPutra]

thank you for your feedback! 0. yes, i decided to start covering javascript frameworks/runtimes with the slowest one 1. yes, go automatically prepares and caches sql statements 2. well, i followed the express "best practices guide" where it says to use async functions 3. i gave it 2 vCPU - github.com/antonputra/tutorials/blob/main/lessons/210/deploy/node-app/2-deployment.yaml#L52 4. noted

[@Nurgul626]

@@AntonPutra 1. Just I thought :), postgres client that you use doesn't make prepared statement by default (you need to handle it yourself), so in this test Go does less job than node.js 2. Ahh, I see, you use express 5. Case dismissed :). Note: the vast majority of projects still use express 4 with no async/promise support. (version 5 got released 16 days ago, after 10 years of development, so the adoption of version 5 is minuscule)

[@SilasDuarte-e9k]

FastAPI, Elysia, Hono vs Go! There are a lot of frameworks that claim to be fast as Go

[@e2e23e]

IMO, if you are using the Go standard library, you also need to use node builtin HTTP server

[@ha5anmukhlis]

you should use a faster nodejs web framework like hyper-express. express is notoriously slow!

[@everyhandletaken]

That's true, but it is a fun real world test, as Express is extremely popular out there too.

[@sama7496]

@@everyhandletaken this is the first time i am hearing about hyper-express

[@skuwamy]

@@sama7496 there's a lot of web framework in js hahahaha.

[@everyhandletaken]

@@sama7496 I haven't heard of Hyper-Express either, just regular old Express, which despite the name, isn't winning any performance awards 😂

[@Gruak7]

You should try a more modern js web framework such as Hono.

[@AntonPutra]

i decided to start covering javascript frameworks/runtimes with the slowest one, next, deno vs. node.js vs bun

[@AlexanderBorshak]

The Node was created to resolve the 10K requests problem. So, it resolves exact this problem )))

[@AntonPutra]

haha

[@j.r.r.tolkien8724]

What a great way to start a "fair" comparison. Literally the slowest node.js framework vs Go's std. Why not use framworks from both?

[@tunexpectedv0id]

The nodejs number suggest you have to rethink your nodejs deployment, at minute 3:02 I think the 56.7% is the total cpu usage in the cluster but I think you need more nodejs instances per node. It is possible to get nodejs reaching 100% cpu (even more than 100% if you account for OS threads doing the IO work, 1.4 cpu cores per nodejs instance is a decent assumption), Go by default will use all the resources much more efficiently without hiccups, nodejs is a bit more tricky to get right.

[@vitor14]

I think the biggest problem is that you are not using clustering, which is something that comes by default in go, it is a little unfair to compare them without worker threads to take better advantage of the available cpu, obviously you are not going to take all the advantages of clustering with 1 or less cpu but in this case I think it would improve seeing that the cpu only reaches 50%.

[@fortuneosho8137]

Spotted a nodejs apologist 😂

[@_Verac]

​An apologist? Nodejs is not a religion that making an argument for it would make one an apologist especially when the argument is purely logical.@@fortuneosho8137

[@elvispalace]

This video basically explain why performance issues is more related with skills issues

[@bjbegui]

@AntonPutra you have an amazing youtube channel

[@Shenepoy]

how much does it cost to run those tests, I would assume not less than 200$ for the entire test

[@skuwamy]

Goodluck! Javascript kiddies are coming....

[@ADP4031]

As a javascript dev, I don't disagree with the results. Even the CPU usage doesn't mean Javascript uses less CPU, it means it didn't managed to use it to its full potential.

[@powerofpavan7710]

​yes

[@TheRanguna]

Go is naturally optimised to scale itself in the same machine. Whilst node is made to be single threaded by nature. Does this setup scale the node instances in the same machine or do we have a single node instance on a 16 vCPU machine? What do you mean "increasing the number of replicas"? Clearly node is not filling the whole cpu.

[@TheRanguna]

My comment got deleted because I pasted a link. Seems like the deployment spec for the node app is assigning 2vCPU to each node process, meaning 1 vCPU is wasted. Which could explain why the CPU usage doesn't go above 60% Meaning that in its current state, this is equivalent to comparing two CPUs, one with a multicore stress test and another one with a single core stress test and concluding that the CPU that was tested with multi core test performs better than the one tested with the single core test. In other words, this test proves that node was indeed made to be single threaded, and if you want more performance on parallel tasks, you can just spin up new instances. Unless I'm reading the deployment manifest wrong.

[@dweblinveltz5035]

@@TheRanguna I believe you are correct; without even seeing the CPU allotments, he did not show any Node.js code that sets up Cluster mode, so... and he mentioned multiple replicas in a cluster, but I don't believe he's scaled them up to match the number of threads Go may be using internally.

[@TheRanguna]

@@dweblinveltz5035 A new video was released with this and other adjustments as well

[@dweblinveltz5035]

@@TheRanguna I watched, very good followup. Making this mistake actually provides great insight for his viewers that may not have been highlighted so well otherwise.

[@laskyb4275]

What do you use for metrics? Prometheus/ grafana? How do you prevent this 2 services from using too much performance of your k8s..i noticed that monitoring sometimes require more performance than my app itself if it's small

[@sanchitwadehra]

Dhanyavad

[@AntonPutra]

my pleasure!

[@adcodes]

Can you do something like 10 separate containers running 10 most popular JS libraries working simultaneously vs only 1 running Go?

[@skuwamy]

what the?🤣🤣🤣

[@adcodes]

@@skuwamy 😅🤪

[@arytiwa4351]

please do it with bunjs and deno 2.0

[@ijazkhan3335]

It was a pretty obvious result. Go is vastly superior in every kind of benchmark that relates to server side. But you did not run NodeJS the way it should be in production.

[@AntonPutra]

well, just released an updated video and fixed some mistakes

[@easy-draw]

Not sure using express in Node vs Go is comparable at all. You could go even slower if using Node with NestJs. 😀

[@Error_00101]

Answer to your Question as to why node is stuck at 60% usage: Node itself (by default) runns applicatins only on 1 cpu core. -> first 50% the other 10% is the kernel and the node environment doing things.

[@pierromojito]

Do a Go vs Bun (JS) benchmark pls

[@AntonPutra]

yes, after deno vs. node.js vs bun

[@stevanmoeladi1058]

Can we see Deno 2 vs Golang?

[@arrow_function]

Do please node vs deno vs bun test. It will be so interesting to see

[@zakimimit]

Can we see: -Django DRF vs Django ninja -Django with celery Vs nodjs or Go This will be fair comparison Thank you for the content

[@pawegraczyk6050]

You are giving 2 virtual cores to nodejs that uses only one, due to the way node stack works. The test should be about how stacks compare in speed when accessing underlying resources fully.

[@AntonPutra]

yes, just released an updated video with improvements

[@ehfoss]

Could you do Python FastAPI v Node.js?

[@hex6651]

Hi! Will you sometime make C++ comparison to GO for example? Seastar framework looking so promising for your benchs

[@SamerLOLOfficial]

I was waiting for this a long time a ago, can you please compare different nodejs frameworks, like expressjs and nest js, and bun and deno and GO? since Last time you compared go with node js you didn't add measures for # of requests. Also why the numbers you have is different from what published on internet?

[@AntonPutra]

next i'll do deno vs. node.js vs bun. when it comes to numbers, i use what i have in my repo for tests. if you think i misconfigured something, please let me know source code - github.com/antonputra/tutorials/tree/main/lessons/210/node-app

[@ulrich-tonmoy]

Python vs Node would be a great comparison both are interpreted And Node vs Deno vs Bun would be a great one too

[@firefox_tn1815]

Why not fastify

[@upcomingweeb136]

Won't still perform as good as Go

[@тимур_атмосферный]

​@@upcomingweeb136That's not the answer

[@baxiry.]

no-optimazed vs no-optimazed. this is fair

[@Balaji-ed3ms]

Deno 2 vs Golang?

[@gabrielmartinez2455]

I want to see C++ vs Go, also C++ vs rust 🎉

[@AntonPutra]

ok 😊

[@ParthaSarathylink]

Hopefully AWS gives you discount for your services

[@rankala]

Elixir or gleam API would be very interesting.

[@AntonPutra]

noted

[@MrChiefTabaka]

I’m curious what software are you using to visualize those values; I mean those charts, is this some kind of dashboard?

[@AntonPutra]

Sure, i'm using open source prometheus to collect metrics and grafana to visualize them

[@lraondua]

fastify vs express would be cool

[@AntonPutra]

ok noted!

[@farSleeper]

Название NodeJS VS что-то. Открываешь код - express. Классика прямо :)

[@Александр-ф9в4ю]

Адепта ноды порвало. Сейчас автор пойдет разбираться в миллиарде нодовых Фреймворков, которые будут лагать все как один даже на фоне го 😢

[@antnzr656]

@@Александр-ф9в4ю та смысл в том что бы если ты берёшь стандартрую либо в го для сравнения, то и для ноды бери так же.

[@Nurgul626]

​@@Александр-ф9в4ю кто бы мог подумать, что чтобы сравнить производительность правильно, нужно разобраться в том что тестируешь

[@dromix_ai]

I noticed in one of your videos that go is able to handle ~20k RPS, here is ~75k RPS. What is the difference between tests? EC2 instances?

[@0xd3c0d3d]

I want to see tests with C# with NativeAOT, it would be something new on yt

[@temperkan3727]

Go: “ I don't make mistakes. I'm not just like the rest of you. I'm stronger. I'm smarter. I'm better. I AM BETTER.”

[@AntonPutra]

😊

[@bopon4090]

Whats the conclusion?? Why is go memory usage going up?

[@AntonPutra]

the standard library does not have rate limiting middleware, and it caches all the requests. eventually, they pile up and exceed the memory limit

[@timelimitexceeded0]

we need comparison between go standard library and bun runtime without any framework

[@chiebidoluchinaemerem5860]

Please which tool do you use for the benchmarks

[@thegrumpydeveloper]

Graviton arm vs x86, also what’s the graphing tool? Also laravel vs node fastify vs go

[@gameloop9861]

Dude, node is called node for a reason. It's single threaded and must be deployed in multiple instances in production with load balancing.

[@AntonPutra]

Yes, I'm debating whether to run 2 replicas with a 1 vCPU limit or 1 replica with 2 vCPUs but limit the Golang concurrency to 1 by setting GOMAXPROCS=1 to make it a 'fair' test. I use an m7a.xlarge instance for each app.

[@ottohavasvolgyi2448]

I wonder why node's data access slowed down so much, while go's didnt at that load

[@ob34915]

Node is single threaded after 9000 req per second the thread will start slowing down its expected

[@AntonPutra]

yes, just released an updated video with improvements

[@mdjpurdon]

I was looking for the container you used to run the test client, `aputra/load-tester` but I don't see it on docker hub or in your github repo. Any chance we could have a look at that?

[@AntonPutra]

sure, i need to refactor it before i can show it to the world 😊

[@JesseLZamora]

Can you do Swift vs. C#?

[@robertodev2002]

Node is not made to grow vertically scalability, but to grow horizontally scalability, at some point if you give it more cpu or ram, Node will not know how to take advantage of it.

[@ggcc3261]

Have you done C++ vs Go benchmarking yet? That would be fun to see

[@Александр-ф9в4ю]

Have you tried to write backend in C++ yet? That would be fun to see

[@ggcc3261]

@@Александр-ф9в4ю yeah, I've recently used Drogon. Haven't you written a backend in C++? 😆