I'm also curious as to how Google came in 5 times cheaper than AWS. I knew intuitively (and through an occasional pricing spot-check) that GCP and Azure could be price competitive enough for certain workloads to be worth considering (assuming a spikey/elastic enough load that cloud makes sense in the first place), but I never saw that extreme a difference.
There is a tool to build around that.
Interestingly, many of the comments are asking for support for GCP, although the response seems to be that Google's tools might be better (enough) than Amazon's that it's not as urgent.
I sent you a mail yesterday. Maybe it went to the spam folder.
To be fair, the machines you specify are pretty tiny and would easily fit into a half-U enclosure. Even assuming only 36U usable (after network, power, etc) per rack, those thousand would fit in 14 racks, a very small fraction of a whole datacenter.
(I did see the e-mail.. I think I have HN emails explicitly bypass spam. I'll check out your site and reply privately later today)
As do I, although there seems to be some kind of mass delusion in effect that causes disbelief of this.
I'm careful not to quote a single multiplier, though I do point out that it's a multiplier, not just merely a percentage (below 100%) premium. I've found that it can vary between 2x-10x, depending on ones specific performance needs and how much one is paying for colo "space" (power).
Part of what I've been realizing is feeding the misconception (if not mass delusion) is that the knowledge on how to price/negotiate commodity servers without getting, shall we say, fleeced, is quite rare. Getting a good deal on the base server and CPU doesn't matter if the bulk of the order is RAM and storage and you're grossly over-paying on those.