[krobertson]

All of those systems introduce convenience at the cost of lockin and they are no longer suitable once you reach a certain scale. Then the problem becomes either the lack of visibility or the cost with the metered services.

For me, a nice service like PubNub would run us nearly $300/day. Compare that to a simple EC2 m1.large at $240/month.

Visibility comes most of all with DBs where you want to manage lower level settings. Things like MySQL configs, disk volumes and their configuration, etc. Good luck tracking your IOPS on a hosted service, or since they're likely cloud based, getting good throughput/latency levels.

Overall though, sysadmins are changing into devops. They're the ones who connect the dots between what the app is doing, what scale it needs to run at, and the systems need to support those. As things grow, need management of all the moving pieces. Then comes monitoring minute aspects of the environment to ensure performance and stability, paging when something goes out of bounds, etc. And as you grow, minor changes or upgrades can have a huge impact or require a lot of roll out, so need to test and benchmark several aspects.

Sysadmins in the small scale may be less important, but they're becoming even more critical as you grow. The nice part is the tools they can leverage are growing so you can do more with fewer hands.