it is good to have cpu metrics, but specifically these two LB are network intensive applications, so network bandwidth are important, if you run the test on the same time, that is not fair, you need to run separately and one machine to other machine even on virtual network, running the test on same machine while running the nginx and envoy is not representing the real scenerio, and cpu is less expensive then memory so I would rather measure memory usage while running tests. Thanks appreciated for this effort, really good setup.
I’m glad you said this. I’ve been on the internet for a long time and watched a lot of videos with benchmarks. I never expected to see one where two different proxies were run on the same server in the same test. Good lessons here for how I’ll improve in the test plan.
Very interesting results across your series. I would like to see your analysis of Apache APISIX. Seems like a very appealing option and they make some very high-performance claims on their home page. Keep up the awesome work!
Hi Anton, awesome benchs! Can i request also tricky comparison of "nginx-unit php app" (i can help with setting up docker image and code) vs nodejs app?
Would you say Envoy would ever be a better web server/reverse proxy than Nginx? Envoy seems better suited to be service mesh or sidecar and also exposes more metrics than Nginx, but like you show here by default Nginx will always be better reverse proxy or webserver.
Well, based on my experience, it was built to be programmatically configured. One main use case is to be part of the service mesh. On the other hand, Linkerd was using it for a while and then dropped it in favor of an in-house built proxy in Rust, which is even more efficient and faster. Also, there are sidecar-less service meshes such as Cilium, which make Envoy obsolete.
🔴 - To support my channel, I’d like to offer Mentorship/On-the-Job Support/Consulting - me@antonputra.com
it is good to have cpu metrics, but specifically these two LB are network intensive applications, so network bandwidth are important, if you run the test on the same time, that is not fair, you need to run separately and one machine to other machine even on virtual network, running the test on same machine while running the nginx and envoy is not representing the real scenerio, and cpu is less expensive then memory so I would rather measure memory usage while running tests. Thanks appreciated for this effort, really good setup.
Thanks, noted
I’m glad you said this.
I’ve been on the internet for a long time and watched a lot of videos with benchmarks.
I never expected to see one where two different proxies were run on the same server in the same test.
Good lessons here for how I’ll improve in the test plan.
I dig all your videos. Suggestions, could you please make one video comparing spring boot and Quarkus? Thanks :) ❤❤❤❤🎉🎉🎉🎉
sure i can =)
@@AntonPutra looking forward for watching it :)
Very interesting results across your series. I would like to see your analysis of Apache APISIX. Seems like a very appealing option and they make some very high-performance claims on their home page. Keep up the awesome work!
Thanks! Sure, will do!
Thank you for bringing this content! this is very insightful
my pleasure!
Thank you for your content. Could you compare Nginx with HAProxy, and Traefik with Kong?
my pleasure, sure I'll put it in my list
I have never heard of Envoy. Now I know. Thank for video!
It is a proxy used in the Istio service mesh. It was used in linkerd as well for a while, but then they decided to build their own based on rust.
@@AntonPutra, it's also the consul connect proxy.
@@MrLayup good to know, thanks
Came here after the Netflix incident... interesting
haha, it's old i'm about to refresh it
@@AntonPutra Yeah would be cool to see....😃
btw cool animation how did u achieve that wonder!
Thanks, after effects and illustrator
Nice video, thanks!
Thanks Fernando!
Hi Anton, awesome benchs!
Can i request also tricky comparison of "nginx-unit php app" (i can help with setting up docker image and code) vs nodejs app?
Thanks, sure if you can send me an example would be great
Need to compare OpenLiteSpeed, sound like it is much faster also supports Http3
Would you say Envoy would ever be a better web server/reverse proxy than Nginx? Envoy seems better suited to be service mesh or sidecar and also exposes more metrics than Nginx, but like you show here by default Nginx will always be better reverse proxy or webserver.
Well, based on my experience, it was built to be programmatically configured. One main use case is to be part of the service mesh. On the other hand, Linkerd was using it for a while and then dropped it in favor of an in-house built proxy in Rust, which is even more efficient and faster. Also, there are sidecar-less service meshes such as Cilium, which make Envoy obsolete.
So if i use gRPC i'd use Envoy
realistically you would choose istio or linkerd and whatever they use under the hood
Nginx vs Caddy ?
The envoy config is confusing af. Nginx’s config dsl is much simpler to reason abt
yeah i know :)