Linode, Slicehost (RIP), the cornucopia of LEB, prgmr, DigitalOcean and friends target hobbyists and small fleets. Their model encourages precious snowflake named machines that aren't disposable and last a while. That model also falls apart beyond on the order of 50-100 nodes, depending on admin competence and documentation. (If you are holding 1,000 machines named after stars or authors or something together with sheer will, consider going disposable and ask me for a beer.)
Amazon is playing a whole different game. Azure is there too. There's a middle, too, where Rackspace Cloud is ending up. They were headed for the Amazon game and seem to have lost momentum. Come to think of it, the middle is littered with those with their eye on competing with Amazon.
With Google's default project quotas, they almost fall in that smaller bucket too (I was surprised by the core quota in particular), but the limits are easy to raise, so they're in the middle somewhere as well.
That's how I'd frame this, not necessarily amateur/pro, but it does sort of break that way. Snowflake/disposable is also the sysadmin/SRE inflection point.
Amazon is playing a whole different game. Azure is there too. There's a middle, too, where Rackspace Cloud is ending up. They were headed for the Amazon game and seem to have lost momentum. Come to think of it, the middle is littered with those with their eye on competing with Amazon.
With Google's default project quotas, they almost fall in that smaller bucket too (I was surprised by the core quota in particular), but the limits are easy to raise, so they're in the middle somewhere as well.
That's how I'd frame this, not necessarily amateur/pro, but it does sort of break that way. Snowflake/disposable is also the sysadmin/SRE inflection point.