As a startup you want to ramp up as quickly as possible to get to the MVP stage and attract investors to help you move into the next stage. Fair enough, but then what?
Typically, many internet startups start building their apps in a public cloud. Just swipe your credit card and you are ready to go. On many occasions you even get credits for free or at a discount, because hyperscalers see your (future) potential and want to get you in early. So, you have no upfront expenses and can go as fast as you want. Using a public cloud is fine to begin with when testing your app. Yet whilst the success of your application grows, it becomes more and more clear that there’s a price to be paid. This often comes in the shape of costs and inflexibility.
When you’ve reached the MVP phase and attracted some investment capital, much comes down to optimization of the platform as well as your organization. As private equity has a time window of five to seven years, one might argue there’s no need to rush. That’s only partly true, in some cases things need to be done fast and done right the first time. This applies in particular for the architecture of your application as well as the selection of the underlying platform, which in most all cases turns out to be a hybrid cloud.
The drawbacks of hyperscalers
The hyperscalers that startups often use to build their applications offer cloud services based on flexible loads. Practically this means that no matter what workloads you put in their public cloud and when you shift them, you pay for having their full capacity available 24/7. Even if you don’t need it. The default per hour price model that hyperscalers use is predominantly based on this. Let’s compare it to public transport. When you only travel a few days per month, it makes sense to pay for a one-way ticket or a return. When you use public transport multiple times per week or only during off-peak hours it proves beneficial to buy a monthly or yearly pass to save money. The other way around, you wouldn’t buy a day pass for the subway when you knew upfront that you’d only needed to make one return trip.
The benefits of hybrid cloud
The infrastructure requirements of an internet application are actually very similar. Applications are hardly ever a homogeneous lump of code. They exist of a number of components, which can roughly be divided into storage (user data), compute (platform related tasks) and bandwidth (communication). You will not need all parts of your application all the time, yet parts of it are fundamental. This baseload exists for example of the website and the product catalogue. Since this capacity is fundamental for your business and needs to be up and running year-round, it is predictable and can be calculated upfront. Run this bit on hosted servers, perhaps even bare metal, through a (multi-)year agreement with a hosting service provider. This will significantly change your cost model.
On top of your baseload you require computing power, storage and networking capacity to complete a number of tasks during the month. These can involve finances, updating the application, sending out newsletters or adding extra resources during peak hours. Most of these types of activities are repetitive and can be mapped and scheduled, so it makes perfect sense to allocate them to a cloud service that you can use at will. Since these are structural tasks, it is advisable to run them in a subscription (private) cloud service that offers you the flexibility you need at a competitive price, taking another chunk out of your fixed costs. Having taken care of most of the infrastructural matters, one bit still is to be resolved. Irregular capacity requirements that occur either planned or unplanned. This is where ad-hoc cloud services come in and how they were initially advertised; to provide additional resources at ‘burst’ events. This is where public cloud providers can be of service when there are no real alternatives. Purchasing of hourly capacity should be limited to an absolute minimum.
Building your application with the future in mind
Keep these things in mind throughout the whole process of building your application as well as maintaining the freedom to drag and drop parts of your application accordingly. Making the efforts to include an abstraction layer from the ‘get go’ will prove very beneficial. An application that uses, for example, container technology will provide ample options to move workloads and functions between infrastructures. As will it show your forward thinking to potential investors.
Evidently, there’s a tension between the ease of use that hyperscalers offer and the long-term freedom that you need to optimize your platform. Advisable is to make your changes gradually and create a roadmap that will support your growth strategy, whilst decreasing your dependance on hyperscalers.
At a certain stage you might want to do more in-house and grow the breadth and depth of your expertise. Therefore, it’s advisable to invest in knowledge. For example, by hiring an infrastructure manager, startups can improve the performance and efficiency of their application. Eventually it may even prove beneficial to hire cloud architects and partner with a cloud provider to build a hybrid cloud.
Contact us to learn more about Leaseweb’s Hybrid Cloud solutions.