Shad Lutz is our Vice President of Information Technology here at Volusion. He's here to tell you some exciting new hosting developments we have on the horizon!
For many months, Volusion engineers have been working on preparations for the migration to Google Cloud Platform. As we wrap the advance work, I wanted to take a few moments to share some additional details with our customers and partners.
Truly, the fastest way to move really large volumes of data — whether it's across town or across the country — is to load it into a moving van and drive it to the new data center. Of course, that doesn’t really work for our business or yours. To safeguard our customers’ storefront availability, we’re taking a more elegant approach, breaking the Google Cloud migration into five key phases:
Phase 1 - DNS Hosting: Volusion hosts DNS for tens of thousands of domains. Because DNS conforms to rigorous standards and works the same everywhere, it makes sense to get this part done first so our customers can begin enjoying the benefits of super-fast name resolution right away. When shoppers visit our merchants’ storefronts, the domain lookup will no longer have to transit all the way to California, but will instead be served by the nearest Google name server, no matter where in the world they’re shopping from.
Phase 2 - Secure Proxy Introduction: The next step is to make available a new, high-security proxy service. This new layer will bring capabilities like HTTP2, full-site SSL, vastly improved filtering of unwanted bot traffic and integrated DDOS protection. Once fully validated, we’ll begin to transition all front-end traffic through this new service, beginning with early adopters.
Phase 3 - Merchant Web Content and Databases: With the new security proxy handling the front-end traffic, we’ll begin migrating individual merchant storefronts and associated databases, starting with small batches of test stores. Once we’re certain the migration tools are working perfectly we’ll transition the early adopters, then incrementally increase the migration volume, ideally completing all migrations before the retail season begins.
Phase 4 - Shared Integration Services: Once the storefront migration is humming along, we’ll begin migrating shared integrations, ensuring their special communication requirements are fully functional before cutting over.
Phase 5 - Platform Management Tools: The last step is to migrate the tools we use to manage the platform. During this phase, we’ll clean up any random services that didn’t fit anywhere else in the plan.
That’s a lot to do in a fairly short time, which is why we’re doing everything possible to reduce risk and minimize impact to our merchants and their shoppers. While large infrastructure projects can be amazingly complex, following these basic guiding principles dramatically improves project outcomes:
90/10 Planning Rule: Before we migrate the first store, we’ll have spent thousands of engineer-hours training, planning, architecting, testing new technologies, updating our tools and processes and consulting with numerous cloud experts on how to best ensure great performance, reliability and — above all — security.
Contain Project Scope: The easiest way to mitigate risk is to simply eliminate all unnecessary changes. A lot of complex projects fail precisely because stakeholders try to squeeze in “just one more thing.” We’re not.
For example, the most difficult infrastructure changes for us to test and validate have always been operating system, web server and database upgrades. So for this migration, we’ll be keeping the same versions our merchants’ storefronts operate on today. We did those upgrades last year, so we don’t feel pressured to include them here.
Real World Testing: The introduction of any new technologies should involve at least a limited proof of concept deployment. It’s a great way to gain basic familiarity and build confidence. But to really know a solution will work as designed, it’s important see how it operates with real traffic, with the kind of chaotic activity that can only be generated by humans. That’s why we’ve been running our own website through the secure proxy layer described in Phase 2 since April. The results have been amazing, far exceeding expectations.
Rollback Plans: Also critical to migration success is ensuring we can easily reverse any given change at the first sign of trouble. It’s so much better to pause, evaluate, repair and repeat, than to just blaze heedlessly ahead to a target completion date. We can always make up for lost time later once the kinks are worked out; regaining the confidence of our customers is much more difficult.
Smaller Bites: Breaking the project down as I’ve described allows us to manage each element in isolation, so if anything does go wrong, we can quickly identify and address the root cause.
Ramping Up Communication: As the actual migration approaches, we’ll be increasing communication via blog posts, direct email and admin panel notifications. The migration schedule is becoming clear and should firm up over the next two weeks.
With experience working through over a dozen major migrations and upgrades since 2012, Volusion’s engineers are well practiced at safely finishing complex infrastructure projects. Although it hasn’t all gone perfectly, we’ve learned a lot along the way, steadily improving execution and minimizing impact by carefully choreographing every activity. I’m confident this project will yield the same great results our customers have come to expect.