Cloud cost models
When you adopt cloud computing you basically switch capital expenditure for a service-based pricing model.
Getting the best deal means getting the cost model which is right for you. With that in mind, here is a guide to what cloud cost models actually mean in practice.
Service-based pricing models.
These days this is probably the most common approach to billing for cloud computing. A cloud vendor will offer a range of cloud services and companies will decide what they need.
They can then pay on a subscription or pay according to use. Many times, companies will use a mixture of both cost models.
When does subscription-based pricing make sense?
Subscription-based pricing models make sense when you can predict usage fairly accurately. Accuracy is very important. If your budget for too little, you will wind up paying per use anyway.
If your budget for too much, you will spend more than necessary. In fact, you might end up spending more than you would have on straightforward pay-per-use. Get it right, however, and you can make very respectable savings.
Subscription-based pricing models also make sense when you want to “lock in” a price. In other words, they protect you from price increases on the vendor side.
Admittedly, they also stop you from taking advantage of price reductions. This is probably less of a risk. Vendors will typically keep their subscription prices lower than their pay-per-use prices.
It’s in their interests to do so. This means that if pay-per-use prices dropped, your vendor might agree to reduce your subscription cost (or increase what they gave in return).
One way to use subscription-based pricing models is to use them for your baseline. In other words, use them for what you know you’re going to need throughout the year.
Then you can top them up with pay-per-use cost models. In fact, many cloud vendors operate “hybrid cost models” in which you start on subscription and then move automatically to pay-per-use if you exceed your subscription allowance.
This can be very convenient, just remember to check how it stacks up against the costs of switching manually to pay-per-use.
When does pay-per-use pricing make sense?
Pay-per-use pricing models make sense for companies that want or need maximum flexibility. They can actually work out more affordable than you might expect.
The trick to making pay-per-use pricing models work effectively is to use reserved instances and/or spot instances. Generally, you want to use a combination of both
Reserved instances
Both Amazon Web Services and Microsoft Azure make it possible to pre-pay for usage packages. In AWS these are called Reserved Instances. In Microsoft Azure, they are called Reserved VM Instances. Most of the time both versions are generally known as RIs.
The key point to take away is that they offer volume discounts without the need to commit to a subscription. RIs can be valid for anything between one and three years. This is ample time for most companies.
Spot instances are basically the exact opposite. They are essentially short-notice, short-term discounts.
Cloud vendors use them to manage their capacity. SMBs can use them to reduce costs. This may require some flexibility on their part, but the savings can be more than worth it.
That said, flexibility cuts both ways. You have no obligation to use the cloud vendor’s services.
The cloud vendor has no obligation to keep their prices at the same level. In practical terms, this is a minor risk.
There is a decent level of competition between cloud vendors. This helps to keep prices in check. It is, however, a possibility and needs to be recognized as such.
User-based pricing models.
User-based cost models are exactly what they sound like. They can also work on a subscription basis or on a pay-as-you-use basis.
The big difference is that you are unlikely to get either RIs or spot instances.
User-based pricing models can work really well for smaller companies. They can blend the stability of subscription-based cost models with the flexibility to scale up and down as required.
User-based pricing models typically work best when companies have limited core staff.
They may, however, add extra staff at peak times. They may also work with freelancers who need access to their systems.
These cost models do typically need to be watched particularly carefully. There tends to come a time when they become financially inefficient. That said, some companies may still prefer to stick with them for their simplicity.
In short
The very smallest SMBs could find that user-based cost models are the most sensible way to go. You eliminate your capital expenditure and replace it with predictable monthly billing.
Larger SMBs are probably going to want to look at service-based pricing models. Subscription-based cost models provide long-term stability. Pay-per-use models offer maximum flexibility.
They can also be very cost-effective if you use RIs and spot instances.
See Also