Why is cloud priced the way it is?
Many people find that cloud pricing seems very complex and different to other IT services, and it often doesn’t fit within the standard budgetary and procurement processes that exist in their organisation. Strategic Blue specialise in understanding the finance and cost optimisation practices of cloud and advising our customers on when best to use them. In this blog, I’m going to do my best to explain why they exist, what they offer and when you should consider bringing them into play.
To understand cloud pricing, many people think you need a PhD, however. when we take a step back and try to understand the motivations behind cloud pricing it suddenly starts to make more sense. I’m going to try to make cloud pricing sound as simple as possible, but I do understand (all too well) that when you get into the numbers this simplicity can quickly be lost. My point, however, is that if you understand the principles, the mechanisms and the reasoning, it becomes a lot easier to model (trust me).
So let’s start with the simplest concept in pricing, supply and demand. Cloud is clearly in high demand therefore the providers need to make sure they have enough spare capacity to handle these new customers. If they can’t take on new customers, they are losing potential revenues and as compute is seen as a commodity the concept of exclusivity being a reason to not have availability falls over. As such, they need to have enough “headroom” to handle the constant new demand.
This constant new demand comes in many flavours, but let’s consider for a moment how many people use the cloud for bursting (short term compute) and for Proof of Concepts (PoCs) to decide if to move forward with a software project that may, or may not, end up on the cloud. Why do people do this on the cloud? Well, it’s largely because they can pay for what they use. So once they’ve done testing or bursting they can switch off and only pay for the time it was online.
This ability to pay as you go (on-demand) is absolutely brilliant, it gives great technology/innovation freedom to test at very low risk. This does, however, create a problem for the vendors, if people are using pay as you go, then they can switch off whenever they like. So when you’re trying to work out how much future capacity you need for new customers, how do you know how much of what you’re currently selling on-demand may suddenly disappear without any notice and you’ve suddenly got huge amounts of spare capacity. Now the chances of this are quite low, but when you consider how many organisations remain on demand even for their long term projects and that the providers have no visibility of what it’s being used for (and therefore how long it may remain), they’ve got a major risk.
So how does this affect pricing? Well, on-demand pricing, although considered well priced versus the market, is actually the cloud vendors most expensive offering. This is because they are holding all of the risk financially. You as a customer are gaining huge flexibility and very little risk, the entire reason it’s attractive to you. So as such the provider charges you for the fact the balance of the risk falls squarely on them. So this explains on-demand pricing in what I hope is a very straightforward way.
Let’s now think about their other pricing, we’ll start with Spot or Preemptible. These are extremely cheap services that have some drawbacks given the vendors’ ability to withdraw the offering or change the pricing based on various factors (which I won’t go into as this is where things get complex). Instead, let’s think about why these exist and why they cost what they do.
We’ve already said that the cloud providers are having to build extra capacity to make sure they have space for increased demand, however, it makes sense that they need to stay quite a bit ahead of this demand, otherwise a spike would mean they would suddenly be at full capacity. This means they have to leave quite a substantial overhead, which would in simple economics be unused/unpaid for stock just sat there. So how do they handle this? They offer you a very cheap price on compute on the proviso that they can take it back at short(ish) notice. This means they can then go sell it on-demand or on another of the pricing models for more money. The basic idea is, it’s better to get some money to cover the costs than none, and they retain the right to take it back when they feel they can make more money on it.
Going back to the concept of risk vs price, in the case of spot/preemptible, you get a very low cost (high discount) because you are taking the risk that it may disappear. While the vendors are receiving less money for it, they do so at relatively low risk because they have the control to take it back and charge it at normal rates as led by demand.
So I’ve said I’m going to keep pricing simple, and already you’ve had to read a lot of text and are worried we’re not even partway through! Don’t worry, I promise I’ll be finished soon, and with that, I’d like to bring us to the final of the three main models for cloud pricing that we’ll discuss: Committed/reserved pricing.
Committed/Reserved pricing is probably the easiest to get our heads around as it is most normal compared to other IT purchases or even day to day services. The concept is this, you tell me how much you want and how long you want it for and I’ll give you a deal. With this, even if you don’t want as much you’ll pay at least the amount you’ve committed to. This is a pretty normal pricing model, just think about your phone contract, it works in the same way. I pay for X amount of minutes, texts and data and even though I never use it I get charged for it. In the same way as my phone contract, if I used more cloud then I’d committed too, I’d pay on-demand pricing for the extra! To make life a bit simpler, the providers offer either 1 year or 3-year commitment discounts and the longer you commit to (the more risk you take and the less on the vendor) the better your price.
There is a concept of upfront payment vs non-upfront payments as well with commitments, but to keep it simple this is just a bit like your car insurance. The more you pay upfront the better your overall price. In simple finance terms, this is just a form of interest on your purchase. If you pay over a longer period (i.e. not upfront) you pay some interest on the balance.
So hopefully this has helped you understand a little more about how the vendors price their clouds. As you can see, there are reasons to use all of the potential offerings dependent on what you’re doing. For workloads that can be interrupted and you don’t’ want to pay a lot for, use spot. For an application that runs in a stable way and will be around for years, use commitments. For quick tests, use on-demand. Simple! Well, maybe not in practice, but that’s where Strategic Blue can help, let us work out how to best use the models created by the vendors to get you the right price for your projects.