20% bump for the most budget-sensitive option is a crazy increase. Sure it's fine if you only have one, but I'm sure there are many folk who have hundreds
$100/hour+ human costs to migrate though right. Imagine downsizing 100 web apps, notifying clients of what's going on, and dealing with random breakages. That's not an hour's work
You can get a GCP micro VPS for free, “forever”. I have a free VPS always running, and run beefier VPSs as-needed for my personal projects. For my needs, GCP and Colab (sometimes Pro, sometimes free) are convenient and time-savers.
You can also get a free Oracle Cloud VPS, and Alibaba provides good cloud value also. Many years ago, I enrolled in Microsoft’s BizSpark program, and if they still do that, check them out. So many good options…
Oracle gives you 10TB of monthly bandwith + a 4 core/24gb ram ARM VM + 1 or 2 (can't remember) 1Core/1GB x86 AMD Epyc core... for free. If you are just looking to get a cloud VM for personal projects, it's by far your best option.
It's a bit weird to say considering it's oracle, but there's no catch and no hidden costs. As long as you don't upgrade to a paid account, there's no way to even mess up and get charged
Is that really the case? I just ran the CLI version of speedtest.net and I got way more than the 4Gbps that oracle advertises. Though I only ran the test on ARM, and I know AMD VMs get much less bandwidth. I also wouldn't be surprised if my results aren't representative of real world sustained usage.
I've priced up DO as an alternative to AWS on several projects and it has never been cheaper once AWS discounts (reservations, spot) are applied correctly. You need to consume a relatively large amount of bandwidth before DO becomes cheaper, and even then a third party CDN is likely to be a better solution
The $5 droplet includes the equivalent of $90 worth of AWS egress, although it is accounted somewhat differently.
An equivalent low usage machine on AWS in us-east (t3a.micro - 1GB / shared CPU) on spot instead costs around $2.16, almost 57% less.
So this is where bandwidth becomes interesting, especially in combination with an external CDN.
Say if you're hosting 100 low volume client sites where each client gets its own instance (sensible security choice), DO you'd be paying $284/mo. more than equivalent AWS config, or you could treat the $284 as a bandwidth allowance, leaving 3.155 TB egress in aggregate across all clients before DO/AWS are breakeven again.
If some/many of those client sites consume a lot of bandwidth, the equation quickly changes, but you need to start looking at edge hit rates of an individual application before attempting to make a meaningful comparison.
The last project I looked at this for included several high bandwidth sites with good edge hit rates on an external CDNs, and AWS still came in significantly cheaper