I don't know... but I got a whole lot of signups after posting on the other thread about this, so I suspect it will have less of an impact than I originally feared.
I intend to use s3 to lighten up to load on my bandwidth limits by hosting the largest static files (media files in my case) there, but plan on remaining with the VPSs for all the rest.
Even though it is small, a large number of small users could only add to the problems others are feeling.
Scaling a database in AWS is difficult, no matter if it is relational or NoSQL. We see boxes with hardly any load get 50% IO wait time and stay there for 20 mins. Try and scale anything when you get that.
http://guru.gg/aether/ubuntu-10-4-amazon-aws-free-tier/
tl;dr The default AMI's for Ubuntu take 15GB which will cost you $0.50 a month. This is a quick hack to get down to 10GB.
You can occasionally get deals for ~$15/year for a *nix VPS, and certainly find something usable for ~$8+/month.
EDIT: I wouldn't run anything in production on these boxes, but they are great for dev machines, VPN/SSH proxies for securing open wifi traffic, remote shell, etc.
It's OpenVZ instead of Xen, but I don't really care; I wouldn't really be taking advantage of Xen anyway. (And they have Xen for a couple bucks a month more.)
I have no idea how its performance compares to Linode, AWS, etc, but stability has been fine so far, after 3 months. I kind of expected to have downtime by now since it's so cheap, but pingdom tells me there's been none.
If you're looking for a cheap VPS so you can run relatively unimportant things (ie. not your startup, unless it's just a side project that isn't bringing you any money yet), it's a good choice.
As for cost, it's pretty cheap - $7.95 for a 512MB Xen.
They have a pretty decent service and the guy running it is also a HNer ( http://news.ycombinator.com/user?id=lsc )
Bottomline, Linode has everything of what I am looking for (as I have seen most of the web interface from work). Easy to operate, fast, etc. Is Prgmr the same?
Since I don't like moving to different servers, I will probably be stuck with my choice for a few years. My mind would have been set to Linode if I wasn't reminded of Prgmr. :P
I do have a serial console/rebooter and you can boot in to a rescue image if you mess up your domain so you can fix things without my help... but it's only accessible via ssh, and authenticated via ssh public key. but yeah, overall, we're cheaper than Linode, but right now in most other areas they have me beat. I'm working on it, and I will catch up in features eventually, but right now it's a price vs features decision.
As for the actual metal we run, every CPU bench I've seen, linode wins (I usually do okay vs. slicehost, though.) this is unlikely to change, just 'cause I don't see CPU as a bottleneck. It's very rare I see high cpu utilization.
The biggest problem right now is that Linode has a system, I believe, for identifying, warning, and limiting heavy disk users. I have no such system. Because of this, I believe, they have a reputation for having better disk speeds than I do. I see this as a big problem for me, and I plan on implementing something similar shortly.