Optimizing Kubernetes Resource Requests/Limits for Cost-Efficiency and Latency / Henning Jacobs

Приглашаем на конференцию Saint HighLoad++ 2024, которая пройдет 24 и 25 июня в Санкт-Петербурге!
Программа, подробности и билеты по ссылке: vk.cc/cuyIqx
--------
--------
HighLoad++ Moscow 2018
Тезисы и презентация:
www.highload.ru/moscow/2018/ab...
Kubernetes has the concept of resource requests and limits. Pods get scheduled on the nodes based on their requests and optionally limited in how much of the resource they can consume. Understanding and optimizing resource requests/limits is crucial both for reducing resource "slack" and ensuring application performance/low-latency. This talk shows our approach to monitoring and optimizing Kubernetes resources for 80+ clusters to achieve cost-efficiency and reducing impact for latency-critical applications. All shown tools are Open Source and can be applied to most Kubernetes deployments.
--------
Нашли ошибку в видео? Пишите нам на [e-mail]

[@antonsmolkov751]

So few offline viewers. But this is a treasure!

[@aperture147]

Thanks for the presentation! I've gained more about 5% requests per sec, at the highest peak is 7 - 10%. It's a huge performance boost for a small company like us!

[@jayntguru]

Great talk. Lots of good information. Thank you!

[@zampogna]

KUDOS for one of the most interesting speech on this argument. Zalando is the best.

[@umangshucool]

Great talk :) 11:55 overcommit to cpu I guess

[@fredsted]

Great talk, very informative, thanks

[@yogeshkad7466]

How are you calculating used CPU/memory per namespace? Can you please paste the command /program here?

[@mehulkulshreshtha6943]

I wish this was divided in chapters..

[@pmwanjohiKing]

14:51

[@dmitryverkhoturov3100]

By the way, the moment you shared has code with an incorrect comment: 100000 is 100μs and not 100ms. I will update the comment in the Kubernetes PR 111554 shortly.