Disclaimer: Tracetest is sponsoring this blogpost. Use observability to reduce time and effort in test creation + troubleshooting by 80%.
Two...
Some comments have been hidden by the post's author - find out more
For further actions, you may consider blocking this person and/or reporting abuse
Nice article. Except for the cost analysis.
Serverless, in the long run, will cost you more, no matter the cloud provider. There are a few reasons for this.
Fast time-to-market often means suboptimal code and we are paying for pure cpu cycles.
All the niceties, autoscaling, built-in monitoring, etc, that come with serverless cost money, so that adds to the price you pay for serverless.
A well-designed kubernetes cluster will beat serverless price-wise for same type of workloads.
I think you're missing the point I wanted to make.
We're not supposed to compare the two technologies point-for-point, but instead, use them where they excel in their own category. Please don't position yourself in the container tribe or serverless camp. Open your mind to the opportunity, and advantage, both can provide.
Based on your use-case, user traffic, and throughput intensity, both can be the better choice. It's up to you to make that choice.
A much smarter man than me once said:
I trust his judgment. 😊
Thanks for reading, I hope you stick around and read one of my next articles. Cheers!
Exactly. E.g. In IoT doing Machine Learning at the Edge (as in Edge Computing) oftentimes Kubernetes or containers in general aren't possible. Serverless is the only lightweight code that can actually run there. Of course increasingly powerful processors & boards are being developed so it may not be an issue to select either in a year from now.
Please don't use TL;DR for a table of content, but for a summary.
Otherwise, it's an awesome article :D
Extremely well written. Thanks for this!
Thanks, Priyansh! I'm glad you liked it.
Awesome article. Great job!!!
It offers the elements to compounds the "Big Picture" which can guide our decisions.
Thank you.
Thanks! I'm glad you liked it. Hope it will help you in your future projects. :)
FYI AWS Lambda can run for 15 minutes not 5.
Aha, yes! This was introduced soon after I wrote the article. Here's the announcement if someone wants to check it out. I'll update the article.
While KOPS is nice to get k8s up and running, I think with new EKS its even faster by using something like ekscli..
Zdravo..
Ćao! 😄
Yeah, EKS is pretty awesome. But, I'd say it's not mature enough yet, to be used at scale. I picked KOPS because of the huge amount of community support it has. And, I personally just love using it (guilty pleasures). 😅
Hope you liked the read, and stick around to check out some of my future content. Cheers!
Agree, self managed k8s is better as you can control more of it.
My comment was more about people getting to know k8s and its super easy to stand up a k8s cluster..
Some thoughts:
Multiple stops on the cloud compute continuum — all with their own merits
Raman Sharma ・ Dec 22 '20 ・ 6 min read