Drupal 7 Site Performance and Advice

I think you'll find it handles the traffic at least as well as a dedicated server, and will give you a buffer so you can take your time tuning the code and then you'll have a choice to make whether to stay with Pantheon or move away.

Worth noting though is that Pantheon runs on the nginx server and not Apache - if your current system is on Apache (which it likely is) then you could have some issues migrating the site. I had a few but nothing I couldn't handle and I'm a noob-level developer, so if you have real developers working on the project they should be able to handle anything that comes up.

Another cloud option is Acquia hosting which may be better once you have your code tuned - I moved from there to Pantheon because my site couldn't handle the load testing. This is because Acquia and Pantheon are set up differently.

Acquia was actually faster in my case up to a certain number of users, but would choke completely at a certain point. Right now I don't have time to tune my code so sticking with Pantheon, but will look at Acquia again later once my site is more efficient.

I can understand your pain, but unfortunately there's no way for us to tell you if your development team is up to snuff or not. The devil is always in the details. Your performance problems may be caused by some small feature with a big impact, or it may be the result of many small things that just add up. It may be caused by the way that your contributed modules are configured, or it might be a bit of custom code that wasn't built to scale.

Your development team should be able to pinpoint most performance issues within a few hours of debugging (though things can get real tricky when the problems only happen on production). But sometimes it can take several days of solid development to do an adequate fix.

The real measure of an experienced dev team is how they communicate with you. And that you can gauge for yourself.

I second the motion to install New Relic. (If you haven't used them before, I'm pretty sure they still have a free trial of the "pro" plan. But yes, paying for it is definitely worth every penny, especially if your website is a core part of your business.)

That server, if properly configured, should be able to handle that load in most cases. Personally, I provide management services to a site that sees ~1200 users (~900 authenticated) during peak hours, and without change in performance according to New Relic. It runs with a 2GB/2vCPU VM for the database (technically, it's three servers in a Percona cluster, but as little as the database actually gets hit it doesn't really matter) and a 1GB/1vCPU VM for Varnish/Nginx/PHP-FPM/Memcache during most hours with a second 1GB/1vCPU webserver automatically deployed if determined necessary by KPIs.

As for your developers making mistakes, I wouldn't necessary call everything that has a negative impact on performance a "mistake". They may have done the right thing at the time of development - maybe even a "best practice". One cannot truly expect developers to build something that can scale to Facebook-size on the first (or even second or third) iteration. The most important thing for all parties involved in the beginning is getting a working product out the door. Easy performance wins should be implemented, of course, but the extra $10K for 10ms of performance boost probably wouldn't be a good investment when a company is at the stage of "it'd be nice if we could get 100 concurrent users".

This is where regular code reviews can pay off whether between team members are using an outside consultant. One of the biggest advantages of hiring an outside consultant to perform a review is a fresh set of eyes can spot something that is easily (and understandably) overlooked by the original developers.