The opinions stated here are my own, not those of my company.
I just read the article Behind the scenes: How we host Ars Technica, part 1. Of course, let me start by thanking Ars and Lee Hutchinson for posting and sharing this info. My first impression is that their technical infrastructure is, well, pretty much what I would have expected:
Yep, that’s the modern evolution of the three-tier architecture in a nutshell, with a load balancer in front, compute, DB, and file storage, caching of course, serverless for tasks, and search. The biggest omission, to my eyes, is DDOS protection, but as a news site with a diverse published corpus, surely they got something. Their IP netblock is Amazon, so maybe it’s AWS Shield?
Of course, the interesting questions and answers are the one that can’t easily be published: What was their per-month capacity (presumably fixed/reserved), utilization over time, and cost on their old infrastructure, and what does that cost look like now, given that the cloud is elastic compared to classic data centers. And when I say cost, I don’t just mean infrastructure, I mean total cost, including labor before and after (data centers and cloud require different roles, expertises, and physical locations).
As far as I can tell (I don’t research this extensively), most companies treat such numbers as trade secrets, and don’t readily publish them (if there are good and useful examples of companies publishing numbers and costs, including a thoughtful treatment of labor costs, please let me know. I don’t have an urgent need to know, I’m just curious). One notable exception is 37signals and DHH, who published the nice article Why we’re leaving the cloud, which has juicy deets like “We’re paying over half a million dollars per year for database (RDS) and search (ES) services from Amazon”. Like the good, diligent (and sometimes saucy) Internet communicator that he is, DHH followed up with We have left the cloud and We stand to save $7m over five years from our cloud exit.
Of course, what is expensive or not is relative. 37signals is a mature company. For startups, time and opportunity cost are often the premium. And as I mentioned above, you can’t ignore labor cost, because tech labor is relatively expensive to most countries’ median. Furthermore, not every company can have the caliber of employee that 37signals can recruit and retain.
Anyways, I’ll be curious what Part 2 covers on their application stack.
UPDATE – Just saw the article Our cloud spend in 2022, fantastic, I’ll need to read that later.
Leave a Reply