Update: Dear Google,

I am flattered and humbled by my evidently high pagerank and freshness rating, but getting my own post as the top hit for search terms relevant to my question 10 minutes after I posted it does not actually help me. Perhaps I will come to appreciate it if the Lazyweb provides some answers. Over to you, Lazyweb.

The Linode I use to run a domain for my friends costs $30/mo. An equivalent storage + bandwidth setup on Amazon S3 would cost < $2/mo, while the equivalent computing usage level (73% avg CPU!) on Amazon EC2 would seem to be about $55/mo.

Allegedly the CPU usage on Linode is not due to a spambot having taken over my Linux box but rather to rngd, which feeds the random-number-generator entropy pool from the (simulated? multiplexed? not sure how this works on a Linode VM) hardware random number generator. I asked the Linode support people about this and their position was "it's normal". But if that went away, my EC2 cost would probably be around $5 rather than $55, so I want to figure out if I should expect that on an EC2 box.

The RNG-related functionality on the Linode that I am committed to providing is:

* ssh access
* SSL access to Squirrelmail
* (presumably trivial) incidental usage, like maybe password salts every now and then?

On and off (consistently on since I recently had evidence of active snooping), I have used "ssh -D" to proxy my personal web browsing via that machine, but the rngd issue had been around long before that and that is optional functionality.

I'm curious if anybody else has been using EC2 in a similar way, what you're actually paying for CPU, and whether rngd has been a problem.

Update: Ah, I'm misunderstanding EC2. It seems that CPU usage isn't pro-rated; it's just hours the machine is up. So EC2 is not a win for general-purpose service hosting yet, and I probably want to just focus on moving to special-purpose hosting.