Live data from Hacker News

Why I recommended ECS instead of Kubernetes to my latest customer

leanercloud.beehiiv.com

121–126 of 126 posts

Re: Why I recommended ECS instead of Kubernetes to my latest customer

#121

Static host stuff on an S3 bucket / static web app. Blob storage account with a table, maybe an on-demand function app. Sub $15/mo to run your thing until you get real demand, yeah. But its not new, the K8S shtick is coming from investors not tech people. And if its coming from the tech people throw them out of the door. Why are you cooking for 8000 people when 6 are coming over? Why are you building a kitchen to coo…

> the K8S shtick is coming from investors not tech people.

Not necessarily. If you need to deal with many containerized apps that are updated and deployed regularly, k8s is a really great tool.

As a rule of thumb, I'd say 20 - yes, and everything in between - up to you.

Re: Why I recommended ECS instead of Kubernetes to my latest customer

#122

Earlier quoted context omitted.

> Why are you cooking for 8000 people when 6 are coming over? Place I worked at had a service running on K8s with, I think, 4 pods, and it got on average one hit every 2-3 seconds during office hours (and virtually none outside those.)

So I'd say the number of pods was appropriate.

You might have to show me the working there because I'm failing to follow how 4 pods is appropriate for a service doing about 15k requests a day.

Re: Why I recommended ECS instead of Kubernetes to my latest customer

#123
post #17

> When looking at the cloud resources, we noticed many On-Demand EC2 instances with relatively low CPU utilization, which can be expected considering they don't have customers yet. As a software consultant myself, I'd probably stop the conversation right there and ask why they are building such a robust distributed system — SQS, SNS, etc — without any customers. Still want to be deployed in AWS? Toss the damn app on…

I’ve been exploring this lately because, honestly, the cloud is total overkill for small startups and hobby projects. Kubernetes has its value even for small scale workloads like that, but it’s still a few steps more than, say, running a Capistrano script to push your code to a small Linux box with a database on a second one. You’ll get really far on minimal resources these days, especially with cheaper ARM boxes tha…

It’s not the first time I’ve written about this. The hyperscalers are pretty much the most expensive way to build a business that isn’t presently hyperscale, and their ecosystems are increasingly optimized for sprawling stacks built on a virtually unlimited number of microservices.

That’s just not a realistic or necessary approach for everyone.

AWS is engineered for excruciatingly detailed billing right down to the moment you’re consuming or releasing capacity, and that’s how they built it. Managing that spend is exhausting.

My business runs on under $200/mo in Linode compute resources and the performance is significantly better than on similarly situated EC2 instances. We were spending that on databases alone with AWS and getting a fraction of the performance.

I make extensive use of “pure” Linode Kubernetes Engine k8s. It’s portable to any other Kubernetes cluster, and it lets me take my stack _anywhere_, even to a rack in the nearest data center willing to rent me space, if I really wanted.

Re: Why I recommended ECS instead of Kubernetes to my latest customer

#124
post #60

Earlier quoted context omitted.

Why would you plan not to have customers? Don't you think the company is able to forecast demand for a new product launch? Disney: We'd like to launch a new streaming service. Consultant: Great! You have no customers right now so you can run it on a singleton EC2 instance until you outgrow that scale! Disney: ...We expect 20 million people to sign up in the first week

> Don't you think the company is able to forecast demand for a new product launch? I'm pretty sure "follow the forecast" is exactly what motivated that post. In other words, the infrastructure is overkill for the initial forecast of customers. They're not working for Disney.

It wasn't really that the infrastructure was overkill, it was that scalable choices weren't made in the first place.

Remember, the comment I replied to said:

> As a software consultant myself, I'd probably stop the conversation right there and ask why they are building such a robust distributed system — SQS, SNS, etc — without any customers. Still want to be deployed in AWS? Toss the damn app on a single EC2 instance...

But in the article, it's pointed out that SQS and SNS would have been better choices at lower costs for low usage:

> When it comes to the application, if I had been involved from scratch, I would have recommended SQS and/or SNS for the message bus, which are free of charge at low utilization.

Basically, this company is in a pickle because they didn't have architecture experts from the beginning, and the development team started writing an application without much thought to areas where SRE and DevOps teams often get involved: scaling and cost optimization.

Which is another way to say that most startups seem to wait too long to hire DevOps/SRE teams because they are roles considered to be "cost centers:" work that is not directly contributing to the money-making business logic.

Re: Why I recommended ECS instead of Kubernetes to my latest customer

#125

Earlier quoted context omitted.

So I'd say the number of pods was appropriate.

You might have to show me the working there because I'm failing to follow how 4 pods is appropriate for a service doing about 15k requests a day.

15k/day is nothing, and 4 pods are almost nothing. You could probably have 2 and nothing would have changed.

Re: Why I recommended ECS instead of Kubernetes to my latest customer

#126

Earlier quoted context omitted.

You might have to show me the working there because I'm failing to follow how 4 pods is appropriate for a service doing about 15k requests a day.

15k/day is nothing, and 4 pods are almost nothing. You could probably have 2 and nothing would have changed.

Or a single 2 core server, double it up if you want HA…
Post reply on HN