Just reading the blog post makes me wonder if they really saved 500k. How long did it take them to build the solution, how many people built it, how much does the new service cost while it’s running? How many ops ours does the new service need?
The blog goes into many details that I doubt that they built it in less than a month. So maybe they will save money over years but they probably lost money while building the new solution
They usually don't make sense optimizing before achieving these levels of expenses. I mean, for a really long time the expenses are so low that it would cost more to engineer these solutions in the first place, and then they suddenly run through the roof exponentially.
But mainly, priorities and "it's not that much in comparison to all else". And perhaps not even tracking the costs properly :)
Or, just have the lambda delete the file or change its storage tier once processing is done.
Or, just use a lambda instead of a proxy.
Or, just use cloudflare.
Many times "why not" is more interesting than "why."
But mainly, priorities and "it's not that much in comparison to all else". And perhaps not even tracking the costs properly :)
They have that much faith that the world’s biggest target will never be hacked nor hire a dishonest admin?