Go Hosting Cloud Infrastructure Upgrade
For over a decade, Go Media has been proud to espouse the virtues of hosting with Amazon Web Services. Also known as AWS, this behemoth lapped every competitor in the space to become the preeminent provider of offsite cloud infrastructure. AWS is largely responsible for keeping Jeff Bezos within the top spots of the financially wealthiest humans on the planet. Early in our industry there was a saying “nobody ever got fired for choosing IBM”. I would imagine that mantle now belongs to AWS when it comes to datacenters. After years of inconsistent service with many predecessors, we breathed a sigh of relief our first year hosting with Amazon and have enjoyed 99.98% uptime ever since, over a decade later. I would have to say, it has been worth the money.
Alas, we are moving away from AWS.
This announcement will not devote much time explaining why we are moving away from AWS specifically. Although there have been speed benchmarks showing cloud competitors being more performant than AWS, we would not make the leap to say AWS infrastructure has anything significant to do with our decision. Most of the underlying hardware across the industry is the same and even much the software is the same. It would be outside of expertise to assert any leading cloud infrastructure provider is “faster” or “better” than another. We can only speak to our experience with our specific use case and our rationale to go in a different direction.
Our migration is really about moving away from Plesk™
Go Media has been a fan of the Plesk hosting control panel long before “the cloud” was a household term. Plesk has given our clients and system administrators easy to use tools for website hosting management. Plesk offers a lot of great features, many of which we will miss as we develop our new hosting infrastructure. Ultimately, Plesk itself was never a problem in terms of what it is built for. The reason we are moving away from Plesk is due to what it was NOT built for and that is scalability. Although Plesk is very smart in terms of resource management, it is really a server management tool built for individual service instances. We have managed to scale it in unusual ways to host hundreds of websites. However, this has primarily been an exercise in what is referred to as vertical scaling. Vertical scaling is essentially stacking more resources onto a single platform to try and get more capacity out of it. Although AWS offers heaps of computing power and additional ways to scale vertically, we began to realize there must be a better way.
Enter ClusterCS and Upcloud
Early on in the history of our firm, we tried our hand at running our own servers for web hosting. Although this is somewhat common in the web development community, we found it to be too stressful and outside of our core competencies to justify being on call 24/7. Datacenter technology has come a long way in the last twenty years so maybe self-hosting wouldn’t be as painful as it used to be. For us, we have come to prefer this division of labor with system admin experts and found it to ultimately be more cost-effective… or at least better for our health. The past few years we have experienced an uptick in stress. It began to feel like we were hitting our limits with our current cloud infrastructure and so our search went in a different direction. It is remarkable just how many new vendors have leaped onto the scene this last decade. After pouring over the leader board and zeroing in on the type of cloud hosting design we desired, we decided to go with a combination of ClusterCS for the control panel service and UpCloud as the datacenter provider. ClusterCS offers a managed platform for the type of cloud architecture we were after: horizontal scaling. This is achieved through a curated combination of leading opensource software such as HAProxy, GlusterFS, MariaDB and CentOS, designed in a way where we can add more computing resources and confidently expect the load balancing capacity to scale in a predictable manner. For a long time it has felt like whatever computing power we threw into our cloud at AWS, only resulted in short-lived gains.
Who is Upcloud? We honestly can’t say Upcloud was ever on our radar as a viable cloud hosting provider. Based in Helsinki, Finland, Upcloud was rarely mentioned in the typical industry forums. What we did hear about UpCloud was all good things and they came highly recommended by the folks at ClusterCS, who manage hundreds of servers. Go Media decided to test drive UpCloud on some pilot projects and we were pleasantly surprised. The console was quick. The servers were ever faster. All the necessities in the cloud dashboard were easy to find and worked as expected. Their team was responsive and, maybe most importantly, they are transparent about their datacenter positions. Because we primarily serve small to medium sized businesses, we prefer to place our servers as close to our client’s business region as possible. We were happy to see UpCloud has partnered with CoreSite to provide their infrastructure in the United States. May we someday find ourselves “fired for choosing UpCloud”? It is impossible to tell. But the provider has struck as as extremely talented, ambitious and well on their way to contending with the behemoths. We feel a little competition for Bezos & Co. is a good thing.
What our Clients Want
Speed. Nowadays it has become more and more clear our clients rarely ever need to use the Plesk control panel. They do not lose sleep over where their website is hosted. Our client’s top issue, when it comes to hosting, has consistently been speed – or lack thereof. An unhappy client is what we lose sleep over. Beyond that, most of our clients have come to expect concierge-like managed hosting service with us, allowing them to “set it and forget it” for the most part. It is our job to never rest on our tech laurels. Our goal with this new hosting infrastructure is performance. It will likely take several months to make the transition, but we feel it will be well worth it in the long run. Major web application providers have proven horizontal scaling is the only way to deliver high availability to the masses. Teaming up with ClusterCS and UpCloud will give us the flexibility and performance we have been searching for. We are applying the lessons we have learned over the years to this new build and are very excited about what awaits on the other side of this migration. If you host with us, please be on the lookout for messages from the support team regarding next steps.
Yours in page load speed,
Web Developer, CTO