It seems like the Wikimedia article has turned more people on to MariaDB - but what about Percona? I'm ignorant about the differences. Does anyone have a good 3-way comparison?
Ah, yes. Did my run in trading. Pay (and experience) is great if you have the stomach for it. I found it soul-sucking and ultimately left. You are correct though, trading is extremely lucrative, not even getting into the other benefits.
Depends on where you're living I suppose. In Chicago, a base of $100k/yr for a systems architect seems fair. In the Bay area and other areas that are more expensive, it will probably be more expensive, but I'm not as familiar with those markets. It certainly depends on the complexity of your product too.
Having a platform to simply deploy code is great, but there seems to be a lot of hate for sysadmins over the past few years.
Heroku and the like are fine for prototyping and small apps, but when you start developing complicated and/or popular apps, at some point you are paying a service tons of money to avoid optimizing at the systems level.
I'm a "devops" sysadmin and sitting with my developer every day makes us both so much more effective. "Hey sysadmin, if 1000 requests to this module in my app happen it breaks, can we figure out why it crashes the server?". To which I reply, "yes developer, after we ran some tests, it looks like your code is so efficient and effective, that there aren't enough available tcp/udp ports to support the number of requests it is able to handle. I can increase them on your default AMI if you like. It will result in a 3000% increase in the number of active users we can sustain on a given instance".
The no-ops position is: Let's just pay for more servers.
The DevOps position is: Let's optimize.
The difference may seem small, but if you are paying $20k a MONTH for your Heroku bills, not to mention supporting 3k servers for your app on AWS (or insert provider x), do you still think hiring a $100k/year sysadmin is such a bad decision?
The devs can still develop a method to quickly deploy code... or just use Asgard, and avoid all of the chaos that can happen if something goes wrong. The sysadmins understand your code and how it is interacting with the server. You don't end up paying tons of money to support a very basic application that has simply never been optimized.
Sysadmins in a devops world do not usually have any control over when builds can be pushed - they assist with optimizing. So if something breaks, the blame falls on devs, not ops.
Unoptimized app (rapgenius example): $20k/mo = $240k/year
Sysadmin: ~$100k+/year and a minor (given devops mentality) increase in deploy time = priceless
Not all cases are the same, but the start-up mentality doesn't usually work at scale.