I think you picked the most expensive dedicated hosting provider in existence. Have a look at SoftLayer, Hetzner, Incero, or one of the other alternatives. Also, at these prices, it's not unreasonable to just order any old Dell server for $600 and put it into a colo rack somewhere.
For reference, the dedicated server I rent for my private projects comes in at €49/month and it has these specs:
Intel® Core™ i7-2600 Quad-Core
16 GB DDR 3 RAM
2 x 3 TB SATA 6 Gb/s HDD; 7200 rpm
Support is extremely fast and competent, I'm really happy. To try and mirror that package with any kind of Amazon (or Rackspace it seems) offering would be prohibitively costly.
No ECC (http://en.wikipedia.org/wiki/ECC_memory) in that configuration. That's a serious no no for running anything but personal/low importance stuff. Random bit flips do happen and will ruin your day.
This is highly unlikely. I'm on a team that administers an EC2 infrastructure with tens of terabytes of RAM, and I've never seen anything that resembles memory errors. Considering the findings of http://www.cs.toronto.edu/~bianca/papers/sigmetrics09.pdf, this would be rather extraordinary.
The whole point of the "cloud" is to use as commodity hardware as possible... accepting that failures happen. If you aren't using multiple regions and servers for deployment, with failover designs in your applications you are doing it wrong.
Run n times, compare results. The idea of the same bit flipping twice is fairly ludicrous. These n times could be on the same machine, or distributed and compared with paxos. Then back everything up a billion times. Safety is extremely hard to guarantee, so the closest approximations will be expensive.
I doubt that would cause an instance to fail entirely in a majority of instances. Linux has it's own memory corruption checking and fixing (on 64-bit words afaik), also the customer can code their own sanity checking or CRC data to detect and eliminate them as much as possible. Also, many file or copy operations have built in corruption checking as well. Of course, on top of that you have your virtualization platform which I'm sure has it's own series of sanity checks and complex hardware handling logic.
Ultimately, it's a business decision of risk-management. Also, if you're in any kind of serious business like banking, there is already regulations around the standards of hardware/software you can use.
Users will never know because EC2 runs on paravirtualized kernels that don't have access to SMBIOS, which would tell us that information.
However, if we can determine what machine models they use (assuming they buy branded hardware), then we will know their chipsets, and from there, we should be able to deduce whether the RAM is ECC or not.
Actually when I was at NetApp we got pretty good at figuring out ECC failures. Just write a program that allocates 1GB array. Fill it with random data and compute a MD5 hash on it (you can do that fairly quickly). Now recompute the hash once every 5 minutes for a month. Run that on 125 instances. (that ends up being 1 Terabit of data btw) which makes a good detector for cosmic rays if it isn't ECC protected :-).
I did look at SoftLayer and it's still not significantly cheaper than AWS. Hetzner doesn't seem to have servers in the US, so I've already ruled them out.
Incero seems cheap, but that's the only thing I've seen about Incero. I consider Incero to be in the bargain hunting on WebHostingTalk category of providers. I'm not entirely comfortable hosting mission critical infrastructure there yet.
There are probably a lot of alternatives that other people know more about than I do. Just to address a concern with Hetzner: connectivity is really good there. Last time I was in the US, I got better roundtrip times from my German server than I got from MediaTemple gs in California (though they also have some very nice features). Of course there is no going around light speed the limitation, but I'd say for a lot of projects a European server +CloudFlare is totally sufficient. In the end, it's all a tradeoff between different factors ;)
An issue with virtually all dedicated server vendors is the tendency to grossly under-equip servers with memory: My Mac Mini has 16GB (a $100 upgrade), and it is simply ridiculous that going above 4GB on a server is considered some extravagant feature the requires significant, expensive upgrades of every other component.
In VPS, memory is everything. Memory is the # of VMs they can spin up, which is directly proportional to the amount of money they're making. What you're ultimately paying for is memory. Yes, some bandwidth, but mostly memory. It's what hotel rooms are for hotels.
If we were talking about VPS servers, then sure, however we're talking about dedicated servers. What you are renting is a physical box (such as Dell R210s) sitting in a rack somewhere.
Is it possible that they might want to save money by giving you wimpy memory, because they feel that with the 32gb they give you, they could be making a lot more with something like 64 micros?
16GB of ECC memory for servers is a lot more expensive than your Mac Mini's memory. I recently spent £250 on 2x 4GB DIMMs for a Dell Poweredge server[1] (ok DDR2 is older and therefore more expensive, but still). The same configuration for my Mac Mini came in at around £50[2]
You paid a significant premium for that older memory (a funny paradox of the industry). As P1esk mentioned, you could have gotten 32GB for around the same price. Straight from Dell, paying their inflated prices, on a new R210 it's $173 to go to 16GB of ECC memory.
I've bought a number of large servers recently (using owned servers at colocation in combination with virtual hosts for DR and geo-sharding), and always equip them with 192GB or more. The pricing of memory is so incredibly low it makes no sense otherwise, yet all of these dedicated server vendors act as if 2GB is the norm and 4GB is the advanced upgrade.
For reference, the dedicated server I rent for my private projects comes in at €49/month and it has these specs:
Support is extremely fast and competent, I'm really happy. To try and mirror that package with any kind of Amazon (or Rackspace it seems) offering would be prohibitively costly.