After a long a deep think, I’ve decided to retire my physical hardware and migrate all of my machines to Amazon Web Services. It wasn’t an easy choice, and I feel like I need to spend some time explaining why, just get some things off my chest.
Reason #1 – Our Hardware is End Of Life
Most of the reasons we’re moving are related to our hardware approaching the end of its useful life. Depending on your perspective, this point may have even been reached long before now.
We run 3 servers, all of which are over 5 or 6 years old. They’ve been fantastic machines and we’ve gotten well over our money’s worth out of all three. Overall they still work without issue, but with their age they’ve started to develop some performance and reliability issues that are becoming a concern.
The obvious answer is to buy new hardware, but doing that introduces more issues and further reasons why we’re switching.
Physical Hardware Introduces Constraints
One of the battles I have long fought internally is what hardware to buy next. Aside from the fact that servers are a fairly large investment (more so with small operations like us), there are a couple of different routes to go.
If you shop for new or used servers, you realize quickly that the number of configuration options can get ridiculous quickly. Along with these options, many of them can’t be changed in the field. This means that you’re essentially locked into decisions you make now (2.5″ vs 3.5″ hard drives are one prime example).
With Storehouse being as small as it is, we’re still working to figure out what balance of storage to CPU works best for us, and exactly what our needs are now and what we anticipate them to be in the future. With these uncertainties, the decisions as to which servers to buy becomes more and more difficult, and the repercussions become more and more pronounced.
Physical Hardware is Really, Really Expensive
One of the drawbacks of buying physical machines for such a small company is the sheer cost of these machines. New servers are instantly out of the question, as even basic machines start several thousand dollars without any additional options selected. Used servers are a very viable option, but new performance issues arise, and newer technologies that would directly benefit us aren’t available on the older platforms.
Not only are servers expensive, network gear is equally expensive. This is where we run into more issues, since with our clustered filesystem network performance is paramount. Being able to purchase switches and network cards that are able to handle our anticipated capacity in some cases costs more than the servers themselves.
Reason #2 – Management
A realization I came to in the past few months was the amount of time I was spending on managing infrastructure. I felt as if more and more of my time was spent troubleshooting why servers weren’t online instead of developing new features of bug fixes for Storehouse.
A lot of the problems we were facing were also nothing new. Tasks like setting up highly available database servers took enormous amounts of time simply because I hadn’t done that before and had to learn as I went. Services like AWS offer real solutions to these problems and make it pretty simple to setup new services and ensure they are redundant and able to scale to meet future needs.
Most (if not all) of our outages were caused by human error. A lot of cases were simply that quirks with the setup of our machines came to light at inopportune times.
Reason #3 – Redundancy
Being able to run services in a redundant manner across different physical platforms is a must now. Hardware fails, and I’ll be the first to say that we’ve been extraordinarily lucky in that we haven’t had any major failures in our run (quickly knocks on a wood table).
Setting up physical stuff to be redundant takes a lot of extra planning, work, and a lot of extra cost in some cases. Letting someone else take care of the nitty gritty details has proven already to be a real benefit.
Large, global services like AWS have a distinct advantage because data and machines can be located around the world and the work of interconnecting them in a highly available manner is already taken care of.
A lot of people are moving to cloud services like AWS, but you don’t see anything from them. A lot of the emotions I’m feeling related to this move are nostalgic. There are a lot of fists that happened in that data center: our first servers, our first revenue, the list goes on and on.
Looking forward though, I’m excited to have even more firsts with AWS. There’s still a lot we’ll need to learn and a lot we’ll need to manage, but it’s relieving to take some of the pressure off by letting them handle most of the low-level configuration and management.
My plan is to eventually return to a physical data center, but as a compliment to our offsite infrastructure rather than our primary platform.