Giving in And Moving To AWS or You Won Jeff Barr, You Won

I’m pretty sure that the year was 2008 and I was flying out of the Indianapolis airport to San Francisco where I ran into Jeff Barr randomly in the terminal. I’d known Jeff thru the blogging world and he was in Indiana on a recruiting trip for Amazon to Purdue. He and I talked briefly and then for the next 9 years I’ve watched the ascendancy of AWS from S3 to what it is today — an all encompassing approach to cloud computing. Honestly AWS today reminds me of IBM in the 1960s / 1970s. It is just that comprehensive.

Now while, I’ve used AWS from time to time, its always been on the path of “on demand storage and computing”. I’ve never even seriously considered it for my primary infrastructure. And, yes, I’ve worked on projects where I’ve used it that way. Still, in my head, AWS was an on demand thing. Well, no more. I’m now plugging away on moving big parts of our infrastructure to AWS and I’m pretty damn sure that all of it will move to AWS.

Why AWS?

We have a strong local ISP that manages our boxes and I haven’t had to physically visit a data center now since 2014 — that’s fantastic. But I had four recent experiences that flipped me over to AWS:

1. The 15 Minute Staging Server. I’d done a personal project and gradually built up a running system. I then wanted to get a staging server up and rather than having to setup up another box from scratch, I realized that an AMI could be made which I could then use to create a new box from. *Poof* 15 minutes later after some changes to Capistrano, I had a staging server. Dang!
2. No More 20 Gig Boot Discs. Our stack is rails / ruby / elixir (not much yet but its where we want to go). By the time you load up everything you need in terms of development tools and code bases across a half dozen git repos, our worker boxes tend to run with about 3 to 4 gigs of free disc space. Add aggressive logging and its pretty easy to run out of disc space. Our hosting company is very strict on boot disc size since large boot discs apparently make restores slow. Well that’s fine but I’m a professional and I rarely need restore. Actually I’ve never needed restore at least (knock on wood) here. And I’m tired of running out of disc space at least once a month. Its now happened roughly 22 times since we’ve been at this hosting company. I’m done.
3. “It Just Rebooted.” We recently had a six machine cluster of boxes all go down at once. Someone mentioned that one of our boxes was down and I ssh’d in to find a 29 minute uptime stat. #$*(#$*#()$# An hour and a half later I still didn’t have a response from our hosting firm as to why. And 8 hours later when I emailed their founder I did get a call from my support technician but the best he had for me was “It just rebooted”. Well that’s just not adequate when I’m spending about 50K /year annually. Particularly when some of our boxes have this:
   uptime
    03:27:18 up 724 days, 22:14, 7 users, load average: 1.79, 1.45, 1.45
   If you’re in devops and you’re not curious why some boxes have 2 years of uptime and others have 29 minutes then you’re doing your job wrong.
4. I Want VM Creation / Low Level Access. Even though we’re on dedicated infrastructure access to low level provisioning info / creation of the VMs isn’t available to me. And I’m tired of this. I’d like to have a better understanding of the hardware I’m paying for. At this point, I’m honestly nostalgic for the days of a rack of 1U boxes — that I understood. If things are going to be virtual then I’m better off with AWS where I at least have the underlying access I need.

So its a brave new world from my perspective. Viva La AWS!

Shameless Plug: We’re hiring!

If you’ve got decent Ruby / Rails / Elixir / AWS skills, we’re hiring here at FI-Navigator.