Go (Golang) Performance Benchmark (gnet vs fiber vs fasthttp vs net/http)

[@AntonPutra]

🔴 To support my channel, I'd like to offer Mentorship/On-the-Job Support/Consulting ([e-mail]) 🍿 Benchmarks: kzbin.info/aero/PLiMWaCMwGJXmcDLvMQeORJ-j_jayKaLVn&si=p-UOaVM_6_SFx52H 👋 AWS is expensive - Infra Support Fund: buymeacoffee.com/antonputra

[@gaby_ops]

👋🏻 GoFiber maintainer here, we have opened an issue in our repo to investigate the crash. Thanks for the awesome video! 💪 We can't see anything in the example code that would cause this, so it must be something internal to Fiber.

[@Akhbash]

AWS should sponsor this channel!

[@AntonPutra]

yeah i'll run out of money soon 😭

[@bjbegui]

@@AntonPutra I'll throw some cash your way soon

[@tandemwarhead]

TLDR: net framework choice does not matter for 99% of the workloads

[@AntonPutra]

true

[@therealgamingmaniac]

Yes, if you're confused because of 2-5% difference at most, just switch to Rust,C,Zig or C++

[@Kirides]

but my router-microbenchmarks tell me otherwise! gin/echo/... are waaaay faster.... yea - people really don't understand their own performance needs 99.9% of the time. Most of us don't build a highly scalable memory-based cache services, but regular "LOB" applications that talk to other services.

[@winfle]

⁠@@Kiridesif you need performance, why use go? ;)

[@dejangegic]

​@@winfleBecause Go is very fast for a garbage collected language. Most of the infrastructure can be built with it actually. Do critical parts still might require some MMM control, but in general, it's great. Especially compared to what's usually used: PHP, JS, Ruby, Python etc.

[@MrBoBrilO]

We need a guide on how to do these tests ourself >:)

[@AntonPutra]

i have all the pieces on my channel from creating an eks cluster to monitoring containers in k8s and source code is under each video

[@jordinario]

Hi Anton, been following your videos for a while now, they are always great! Hopefully you can keep up with it for the longest time! Cheers from Brazil

[@Qrzychu92]

as always, great video. Maybe small feedback, could you show also actual MBs of memory in graph? It's hard to compare the values between different videos, and I don't expect you to publish every possible permutation of technologies :)

[@AntonPutra]

got it, i'll convert it to actual usage next time

[@snowleo777]

Круто, вы делаете интересные и актуальные тесты. Всегда с большим удовольствием смотрю ваши видео.

[@nabilpatel970]

JavaScript backend libraries. Such as NestJs, express, fastify etc please

[@roccociccone597]

Lol imagine using JS for backend services.

[@dermuschelschluerfer]

@@roccociccone597 sounds like guaranteed 100ms latency form the start

[@mujibulhaquetanim]

@@roccociccone597 what lol! it is being used widely.

[@trustlang_ua]

idea for next video: nest.js with fastify nest.js with express

[@Figurum]

vs bun vs go

[@eduardoolat]

Your videos are gold! Can you please compare SQLite vs rqlite vs libsql (turso)?

[@AntonPutra]

thanks! added

[@TheHTMLCode]

Awesome as always. I’d love to see Go vs PHP8 vs PHP8 using Swoole

[@AntonPutra]

thanks! i'll do php soon

[@mohit84604]

Event loop cannot be fast when you have computation in routes,and you are not offloading that thing to another go routine.

[@AntonPutra]

got it thanks!

[@ievgenmajor3301]

Good stuff. Good stuff.

[@AntonPutra]

thank you!

[@dimasarestu2043]

pls benchmark mysql 8/8.4/9 its worth it for upgrade or not

[@AntonPutra]

ok added

[@roganl]

Anton, another beautiful comparison - that these tool authors should be tracking independently. I am concerned that you may be polluting your results if you are running these tests concurrently against the same shared postgres instance - as that's a shared resource that will cross contaminate the results

[@UwU-f2a]

maybe test using a postgresql driver that has pool size setting, the driver must be the same for all the framework here to make it apple to apple, like pgx for example

[@AntonPutra]

yes i use exactly the same driver and not only that, i use exactly the same method pool size setting - github.com/antonputra/tutorials/blob/main/lessons/229/myapp/db/db.go#L13-L14 db insert - github.com/antonputra/tutorials/blob/main/lessons/229/myapp/device/device.go#L27

[@shm236]

Gnet seems interesting. I wonder why their memory useage was so much less in the first test.

[@marianivanov6431]

can you show who is win with image sometimes i cant understand which tech is best

[@AntonPutra]

stdlib is great for low latency, fasthttp for throughput

[@davidgazda4585]

Create please rank table of language+framework with its results

[@papryk1159]

your videos are great man, binge watching them all the time! regarding another benchmark - what about some storage solutions? Rook vs OpenEBS?

[@ListenSomething]

Thank you for the comparison. I just want to tell you can increase volume a bit. It sounds very low.

[@mikolajsemeniuk8574]

Great video !

[@carlhandy]

Always wondered about how much this is costing you. Maybe instead of AWS use Digital Ocean, Linode or some cheaper VPS alternative.

[@AntonPutra]

yeah it is costly

[@jakubklecki2963]

Okay so I'm just going to stick to stdlib, if I ever get 10k requests/sec (very unlikely) only then I will start thinking about other options

[@AntonPutra]

fair but think about middleware as well

[@Nick-yd3rc]

Yes and you need to consider your actual payloads and latency.

[@cheatoffchannel6320]

Python FastAPI please 🙏, it’s gonna be surprisingly fast.

[@gaby_ops]

No, it won't be compared to Golang

[@lachlanm9304]

framework that is never talked about is golang iris, would like to see that compares too all of these ones

[@fusionfuryx]

Can you benchmark FastAPI vs Go stdlib?

[@AntonPutra]

yes but why stdlib and not fasthttp

[@SilasDuarte-e9k]

​@@AntonPutra I believe that in production people tend to use stdlib in Go, and some people believe that FastAPI can be an alternative to Go.

[@cheatoffchannel6320]

Yeah would be great to know how FastAPI stands against any compiled language frameworks

[@roccociccone597]

@@SilasDuarte-e9k spoiler: it can’t. We literally moved from FastAPI to Go.

[@SilasDuarte-e9k]

@@roccociccone597 I know, but some people need to see to believe 😅 I prefer Go over Python 9/10 days

[@w1png]

Any plans for minikube tutorial part 4? I know it didnt get a lot of views but its such a great tutorial series

[@AntonPutra]

i may do more tutorials soon

[@shinebayar]

Wow framework claims that it's fast, but actually is fast? That's new.

[@martinhotmann7868]

Yeeeeeeah :D Finally .. love you so much! (◕‿◕)

[@AntonPutra]

will now compare with other fastest frameworks

[@amig0842]

@@AntonPutra gnet and zzz (zig) 😋😋😋

[@brianarnold9149]

Please compare Flask vs. Django vs. FastAPI And also Go's Stdlib vs. Python's FastAPI

[@Monkeydew1o2]

Please do a higher level golang framework comparison. Like gin versus Chi

[@AntonPutra]

ok

[@FilipeASantana]

Top heim, dublado em português Br. Aí sim eu dou valor.

[@trap7369]

a dublagem é feature nova do youtube kkk, eu achei meio robótico, se não achou?

[@seringturu]

What did you use to benchmark and making that graph?

[@AntonPutra]

prometheus + grafana and i have a client that instrumented with metrics as well

[@snatvb]

once again confirming that you are more likely to get stuck in a database than in the capabilities of your application :)

[@awksedgreep]

Fastest Go web framework: Phoenix ;) Wait, you should totally test that theory. haha

[@101kawsar]

stdlib 😎💪

[@AntonPutra]

yes it's great

[@riskikukuh1293]

Hi Anton, can i run your test in my local machine ? Can you make video to tell us how to run your code / benchmark, thanks a lot

[@AntonPutra]

sure you can source code in the description, i may include compose files in the future

[@riskikukuh1293]

@AntonPutra i mean, you can explain how to run your benchmark 😅

[@zhifong-gan]

can u also test Hertz framework?

[@AntonPutra]

yes i just forgot about it

[@MusKel]

When Kafka vs Nats?

[@himmetavsar2168]

Waiting for this one as well

[@AntonPutra]

soon i promise

[@fprotimaru1944]

@@AntonPutra this would be one of the best videos you ever made:)

[@Nick-yd3rc]

NATS is substantially slower, I’d rather watch Pulsar vs Rabbit vs Artemis. On various message sizes. It gets only interesting if you parameterize your message size. But all of them work best in clusters to serve concurrent workloads.

[@carlos-rian]

The same for python please ☝🏼

[@orafaelgf]

great video, tks.

[@AntonPutra]

thanks!

[@vasilekx8]

What about Gin and Chi?

[@AntonPutra]

they are not optimized for performance

[@zezohassam]

Can you do spring boot vs Django

[@UwU-f2a]

obviously springboot win (JIT compilation, better runtime and garbage collector, better features). Django is pure python, no binding with performant library written in other languages, it doesnt even beat nodejs

[@AntonPutra]

yes looks like a fair comparison from a functionality perspective

[@mmadevgame]

@@UwU-f2a Sometimes it's just nice to see how fast or slow something is compared to something else

[@Nick-yd3rc]

Define Spring Boot. The defaults are fine but there’s a lot of knobs to tune it to your actual workloads. How much RAM, what stack size, blocking, virtual threads or reactive controllers and database?

[@Grahamaan27]

Imo, if a container is getting more than 20k req/s you should start scaling earlier

[@AntonPutra]

true

[@fedot_compot180]

What about bun router?

[@Teo-t9u]

Prisma vs drizzle 😊

[@AntonPutra]

ok will do

[@IlyaShaforostoff]

@@AntonPutra please add sequelize to this comparison )

[@AliResool-k8z]

Where is go gin?

[@BarakaAndrew]

It wouldn't make sense since gin is a wrapper for stdlib, the performance would be the same as stdlib. Same goes for echo. They are all just stdlib with some fancy stuff built on top such as json parsing, context e.t.c

[@AntonPutra]

yeah gin is on the same level as stdlib

[@MilanRodd]

You forgot echo again ((

[@AntonPutra]

i didn't want to disappoint you lol

[@azulamazigh2789]

We demand HAproxy vs NGINX

[@AntonPutra]

haha ok

[@azulamazigh2789]

@@AntonPutra you are the best

[@sumitftr]

Please make: CouchDB vs MongoDB vs DynamoDB vs ScyllaDB vs FireStore

[@nana.bxffour]

Ummm how much does he earn? Those are big boxes

[@AntonPutra]

yeah

[@wwloyd]

std lib is the way

[@AntonPutra]

yes

[@maximtheconqueror]

And... Rust is faster!!!

[@AntonPutra]

rust is coming, will see

[@trap7369]

@@AntonPutra waiting for more rust benchmarks, great job!

[@markxavior]

You should gave included gin

[@AntonPutra]

will do next time

[@AchwaqKhalid]

You forgot the following Frameworks: - Gin - Chi - Beego - Iris - Echo - ReveL Please do include these in your future video/s because Golang is apparently on high demand nowadays 👨🏻‍💻

[@AntonPutra]

ok i'll include some of them

[@revenity7543]

GNet is not actually standard compliant

[@AntonPutra]

it's not

[@SilasDuarte-e9k]

But it's Go

[@revenity7543]

@@SilasDuarte-e9k It's not a valid HTTP framework

[@revenity7543]

@@AntonPutra Are u agreeing with me or no

[@michelvandermeiren8661]

So std is as good as fiber ? Soo many people say fiber is way faster

[@AntonPutra]

It is faster at some point and has higher throughput

[@jesusruiz4073]

Yes, fiber is way faster (see first test), but in typical database applications, it does not matter. A simple example: - HTTP handling time with Std is 1ms, and with fiber is 10x faster: 0.1ms - Business logic and DB operations: 99ms (it is the same independent of the framework). Total times: Std: 100ms, fiber: 99.9ms And normally, the difference is even less. However, when the time elapsed per request is mostly HTTP handling, the difference is apparent.

[@news3951]

Hello Brother Anton Putra 🥰🥰, Can you compare AWS S3, Cloudflare R2, azure blob storage , backblaze b2 storage and Google Storage ? .

[@AntonPutra]

ok i'll think about it