I'm not completely sure that I agree. While this might be great for small startups, once you get past the growth of needing more than a few dozen servers, your cloud and other computing costs are going to be exponentially higher than what they would if you had actually purchased the hardware from your favorite hardware vendor, hired a competent sysadmin, and managed the hardware (and storage/network) in-house.
Currently, at $DAYJOB, our database is growing by about 1GB/day, and our storage is growing about 4GB/day -- in a downward housing market. If/when the housing market kicks up, I expect to see those double, if not triple. Having someone in house who can keep the right people informed of the risk (of running out of space, of buying the wrong equipment) who is focused on keeping of the needs of the company balanced with the needs of the software is critical -- do you honestly think your cloud provider cares if your customers can't access their data during an outage? If you're a small fish...probably not...
I agree with you but I'll go even further and say this guy has no idea what he's talking about. As someone whose been in charge of a fairly small cloud outlay (8 virtual servers, 2 of which go off at night) I can tell you the cloud introduces its own set of challenges.
So there's less hardware replacement and more DNS configuration but none the less there's still a job for system administration.
In fact, I'd say the role of administration becomes even more significant because the cloud industry is so new. Hardware has, in many ways, become pretty rock solid in the last decade. We're to the point where a hard drive in a RAID setup can go out and the users won't even notice.
The same can't be true for things like SQL Server Federation and caching servers which used to be the exclusive domain of large web sites but are increasingly being used by smaller companies who don't have a whole staff to dedicate to their upkeep.
(One Caveat: If cloud models like Microsoft's Azure take over this could change. By that I mean clouds that are literally frameworks in themselves and not virtual machines)
You have to make sure you use the right tools. If you re-invent the wheel a dozen times, managing cloud deployments will be a nightmare. Tools such as Rightscale or Scalr.net are well with the money.
I manage 2 dozen Ec2 instances of various types as a side task to my regular development job. I script everything. When things are scripted they are consistently reproducible and fast. Humans make mistakes (including me).
I would prefer there to be a dedicated SysAdmin but there just wouldn't be a full time job for him to fill.
Currently, at $DAYJOB, our database is growing by about 1GB/day, and our storage is growing about 4GB/day -- in a downward housing market. If/when the housing market kicks up, I expect to see those double, if not triple. Having someone in house who can keep the right people informed of the risk (of running out of space, of buying the wrong equipment) who is focused on keeping of the needs of the company balanced with the needs of the software is critical -- do you honestly think your cloud provider cares if your customers can't access their data during an outage? If you're a small fish...probably not...
EDIT: Spelling/grammar change.