Has parse's shutdown made you re-think about PaaS solutions? Do you think one should trust them forever?
I much rather run my own systems and manage them using DigitalOcean. I have a setup with a master server and the sub servers download and run a SH script from the master server once a day that does things like updates, setup environment changes made and other things.
PaaS solutions would be like AWS Lambda, Google App Engine (not Compute Engine), Heroku, etc where you don't worry about the infrastructure and only deploy your app while the PaaS solution takes care of provisioning, autoscaling, etc.
I've been burned by such solutions in the past, either they change something and it ruins your application or they shutdown and you can't easily move to something else.
I'll stick to IaaS (Infrastructure as a Service) for now where I have full control over the VMs and can move around very easily between providers.
I prefer to choose a PaaS or SaaS that I can "easily" replace if something goes wrong (ie: they shut down)
Hosting companies? Dime a dozen. I host with Rackspace, but I could move to AWS or DigitalOcean or any number of hosting companies. Linux is Linux.
Same with the database. I am and probably always will be a huge fan of SQL. MySQL can run anywhere - again, at Rackspace, AWS, etc... Both of these (hosting and database) if all else fails, will even run on in house servers. With a little work, I don't even need MySQL - it could be MSSQL or Postgres or Oracle or ?
Some systems are out of convenience. I love Cloudinary - I haven't found a suitable alternative to it but if it went away; it just processes images and videos. It'll be a burden to not have it, but it's not mission critical.
Search? I use Algolia. Would it suck if the website doesn't have a search feature? Sure. But it won't take down the site if one day they close their doors and - with a bit of code, could be replaced. They make it easy to search and maintain the index, so I use it. But I could of very well have written some scripts to search the DB and not need another service.
Everything will come to pass eventually. Recognize your potential points of failure and have a plan or even just an idea of what you might do, if that service went away.
If Cloud hosting is considered PaaS yes, in my company we are moving forward into the google cloud :)
Mario Giambanco
Director of User Experience Development
Anthony Rossbach
I develop large scale projects for companies and myself. I also own and run NodeHost, a web hosting company.
I don't think I did a good job explaining, so here it goes. The problem with PaaS is the testing environment is entirely on that provider, so is the live and debugging environments and for me it adds a lot of extra work to make applications work as you intended if you are running local when building and not running a second development version.
A major downside is cost, PaaS solutions for the kind of work I do with large scale is WAY more expensive then running a few servers or DigitalOcean droplets for a lot less. Looking at Heroku to move my single largest application would cost me around $1700 a month, and at this time on DigitalOcean droplets it only costs $70 a month.
I also run a lot of custom solutions like a custom firewall thats database connected so when one server finds a IP to block it will report to the others so they can also block. (today it's added over 200 IP's alone) I run a web hosting company so websites get constant attacks but we never have problems with brute force logins because even with NO Wordpress plugins our system monitors logs for failed login attempts (302 returns to the login page wp-login.php in the logs is equal to a failed login) so we can do a lot more.