By Zack Rosen, CEO & Co-Founder, Pantheon
We’ve all been trained to think of “The Cloud” as generic virtual machines on demand à la EC2. At Pantheon we believe this VM-centric compute model will be supplanted by a simpler, vastly more efficient model of cloud computing: containers running on bare-metal infrastructure provisioned via API.
Last Thursday I had the privilege to present on stage at GigaOm Structure with Rackspace President Taylor Rhodes as they unveiled their OnMetal product, which is designed from the ground up for this new vision of Infrastructure-as-a-Service (my bit is around 9:48):
Containers on bare-metal are orders of magnitude more efficient than VMs
We run over 70,000 websites, some doing more than 100 million page views a day. Pantheon can scale you from a tiny site with no traffic to “Internet Famous” in seconds.
But we also provide development and testing environments for our customers. So if you add up all of the environments we operate, the grand total is 250,000 custom Drupal and WordPress sites, each one security and resource isolated from one another.
If we were to do this the hosting architecture way with Virtual Machines, that would mean running over 250,000 VMs!
Could we afford 250,000 VMs? No.
Could we operate 250,000 VMs? No.
Instead, we adopted the bare-metal and container compute model and have built one giant unified platform, running all 250,000 environments with hundreds of thousands of orchestrated containers, serving billions of page views a month. This unified platform gives us a huge operational advantage versus the VM-centric hosting architecture.
For example, two Pantheon engineers were able to patch all of our sites in approximately 3.5 hours of engineering effort, less than 12 hours after the Heartbleed bug was first announced. Our entire infrastructure operations team consists of two full-time engineers. Our team of 13 engineers deploys an average of six improvements across all of our sites every single day.
This bare-metal and container model has enabled us to be an order of magnitude more efficient to operate our infrastructure versus our competition who are using VM-based infrastructure, which helps us develop our product an order of magnitude faster. We pour this efficiency back into our product, and that’s the root of why developers prefer our platform.
The future of Infrastructure-as-a-Service
With a small engineering team relative to our competitors, we think very hard about the technical bets we are making as a company. The bet we made in 2010 on the container and bare-metal infrastructure model has turned out to be a “bet the company” kind of bet. It’s paying off hugely.
After proving this model for our customers, we now have so much conviction about this compute model that we think it’s the obvious future of cloud computing. It’s simply vastly more efficient. Interestingly, the efficiencies we’ve seen first hand at Pantheon are known to Google, which has been running on containers for almost a decade.
We’re still in the very early days of cloud computing. What we’ve seen so far, VM-based Infrastructure-as-a-Service has been powerful, but it’s just a stepping stone to the real value. The best way to make strategic bets on cloud technology is to roll up your sleeves and try this new technology. The good news is much of it is open source, and the cloud model is cheap to try. You owe it to yourself, your team, and your company to get these bets right.
This post originally appeared on Pantheon’s blog here.