This!

This should be emphasised above all other concerns - AWS is absolutely the WORST choice to go for if the company is already struggling with money or not (yet) making any money. It's literally a hole for burning money.

You simply don't go for AWS until you have a superb business plan how to make significantly more money off of your service than Amazon does.

> We moved the platform and every photo to Amazon Web Services (AWS), the industry leader in cloud computing, and modernized its technology along the way. As a result, pages are already 20% faster and photos load 30% more quickly. Platform outages, including Pandas, are way down. Flickr continues to get faster and more stable, and important new features are being built once again.

SmugMug did it primarily to improve stability and performance. They point out that Flickr hosts tens of billions of photos with 100M registered users and hundreds of thousands of paying customers across the globe. I see the need to use AWS (or any cloud provider for that matter) as a necessity to run a paid web service cost effectively at this scale.

It's a bit of a chicken-and-the-egg problem, but not really. They still needed a solid business plan how to earn money from Flickr before moving to AWS (or any other expensive global cloud) - this kind of looks like they've thought of asking for money AFTER the move.

Hope the 30% increase in performance was worth it...

> They still needed a solid business plan how to earn money from Flickr...

SmugMug, by their own admission, bought Flickr to help keep it alive. Currently, they're not even looking to make money but reduce the operating costs. I'm sure they didn't take the decision to move to AWS lightly, at all.

> ...this kind of looks like they've thought of asking for money AFTER the move.

May be you're right that it might have been a mistake, but the fact that SmugMug are a heavy AWS user means it might have made a lot of sense to leverage not only the engineering and operational expertise but also the heavily discounted rates they must have negotiated with AWS?

Bandwidth and storage is likely still killing them. Should've put it in Backblaze B2 and paid Cloudflare for CDN.

Would probably cut their image hosting costs like 70%, maybe more even.

That's like saying "if a car manufacturer is really large, the only way to make it work is to outsource its car manufacturing". I guess so, if they forgot how their core business works. Seems like Flickr forgot how to website.

I was wondering how long it would take for the HN experts to show up.

Indeed. I fix AWS bills for a living, and I wouldn’t dream of saying whether this was a smart move or not without a careful look at the infrastructure, the financial model of the business, and a series of discussions with the engineering and finance teams.

Somehow HN experts can bypass all of that and do what I do full time with basically no information about the company in question. I really should hire a few of them.

Well, now we in a need for experts who fix AWS billing for a living.

lmaoo

If you're spending far more money maintaining technical debt-ridden on-premises solutions, then moving to AWS most certainly is the way to go.

What is the way to go?

For a large-storage, network-heavy content, optimizing the storage and network delivery costs. Colocation for cheaper storage, peering and direct ISP contracts for the network.

I was going to say if they are using a low-cost CDN it's not that bad.

My guess would be their main costs are from hosting all those high-res images, not anything they are doing with EC2 instances. So who cares if they pay a little more for their servers?

But I did a lookup - and live.staticflickr.com is using CloudFront for their CDN. That means they're paying a 5X premium just for image hosting. No need to have the CEO make a blog post acting like he's the Jimmy Wales of image hosting... just switch to a cheap CDN already.