How are we going to use aws ( or azure/google)

Infrastructure
#1

hi all, I’m trying to get a grasp of what it is exactly that you plan to run on aws ?
I’m not sure if it is going to virtual machines or the services

services like mariadb/webservers/gitlab/awx/idm

In my experience as an infrastructure architect, you are going to be in for a financial nightmare if you are going to run virtual machines hosted by a cloud provider like aws/azure/google. There are many much less expensive hosting providers that provide that service.

if you look up the pricing of their virtual machines hosting you will find that after a year you could have paid for a high end server stored in a rack at your local hosting provider.

or to put it short, renting a virtual machine in the aws/azure/google cloud is insanely expensive.

it’s a lot cheaper to rent some real iron or a virtual provided server at a provider that is not AWS/Google/Azure

you can always create you own HA by renting multiple cheap vps at several hosting providers, then connecting them with vpn (like wireguard)

Rob

4 Likes
#2

Yeah and that ain’t the half of it. The bandwidth costs will be massive!

#3

Hey @RobVerduijn!

I work in AWS every day, and I hear you. In my experience, the costs are much better when you have someone who can bridge business need with spending and take advantage of things like AWS Reserved Instances or Spending Plans.

Right now the infrastructure team is working on AWS with vendor agnostic methods and tooling.

One of the reasons for that is that the project is still in a startup phase, and it’s a lot easier at this juncture to build on existing, high-scale, high-availability, secure platforms than to start out building our own bare metal infrastructure. Right now there’s a big focus on getting applications and architecture running, and a keen interest in doing that in a portable manner. This is all driven by a goal to try to provide a build that can reliably replace CentOS by the time CentOS 8 is EOL.

(AWS is known to provide free transit/object storage to large-scale FOSS mirrors. For example there are free, CentOS-supported CentOS mirrors in (almost?) every region on AWS for any EC2 instances to use. So hopefully in time Rocky Linux might have that advantage because CDNs and repos are notoriously bandwidth hungry as @eva2000 pointed out!)

With that said, there are some conversations about some bare metal infrastructure that are happening, so these concerns are not being ignored. If you’d like to participate in those in-the-moment, you can join in on Slack, which you can join at the link on the website.

I’m not trying to shut down this dialogue, I just wanted to bring some extra context. :slight_smile:

3 Likes
#4

I didn’t know that. That would certainly help too. I’d just be wary of engineering it so you ended up locked in with AWS.

For alternatives also check out former packet.net now called Equinix Metal too https://metal.equinix.com/product/servers/ some of the customers https://metal.equinix.com/customers/

API & Docs https://metal.equinix.com/developers

They also have reserve/spot pricing too https://metal.equinix.com/product/pricing/

bandwidth costs are way cheaper than AWS

How does bandwidth pricing work?
We offer premium bandwidth at prices that allow you to scale. Egress bandwidth is just 0.05/GB globally by default. Bulk bandwidth rates can be as low as .005/GB, and require commitment.

Or for object storage, using Wasabi and then Cloudflare for CDN/security/firewall and Cloudflare’s Bandwidth Alliance allows for free bandwidth between Wasabi and Cloudflare https://www.cloudflare.com/bandwidth-alliance/

#5

hey Eva,

appreciate the concern. As one of the infrastructure managers for Rocky Linux, is high on my list to make sure we avoid vendor lock in, and deploy things in a way that allow us to move wherever we need to, whenever we need to

2 Likes
#6

Hi,

first
…already in the rockylinux slack, but it’s kinda chaotic, and hard to follow when you come back online from a good nights rest.
Also on rockylinux irc … same problem

The forums are better organized, and you can properly read up on the entire discussion regarding a subject.

as to hardware virtual or real.
I read the link regarding the methods and tooling, it suggests we use the services on aws not the virtualization, this is good.
and vendor agnostic which is even better.

I’ll definitely look into this Rocky's Infrastructure Platform

Did you get this from a sales flyer ?
‘bridge business need with spending and take advantage of things like AWS Reserved Instances or Spending Plans’
I’m gonna remember that one… sounds great.
How about we change it to…
bridge business need with spending and take advantage of platform agnostic services offered by major cloud vendors.

Rob

#7

That little line came from my own experiences or those I’ve witnessed. I have seen first hand how easy it is to not pay attention to costs in PaaS/IaaS environments. Giving everyone an account factory, for example, can be a costly game! :wink:

I absolutely cheer on vendor agnosticism alongside you. I’m all for something like Ansible, Terraform, Puppet, et al on top of the major players. When called for I will at least use vendor-specific services that are replaceable, like AWS Secrets Manager replaceable with something like Azure Key Vault or Hashicorp Vault. :slight_smile: