The cost of cloud computing is, to be sure, very much of a “pay you go” model. You pay per use and you pay for everything. Cloud providers like Amazon, Microsoft, IBM and Google have spent billions to build out massive data centers the size of football stadiums and they aren’t giving that away on the cheap.
It’s remarkable how many people forget this, or don’t think about the cost. Surveys have found many companies leave the cloud after getting a massive bill because they failed to accurately predict usage fees. At the least, firms should plug in their usage variables to a cloud cost calculator.
The major providers are all offering "pay as you go" access to virtual machines running on Linux and Windows, so the basic features aren’t that different from one provider to the next. And their fees are comparable. It’s the other intangibles that can make your decision for you.
As companies make greater use of the cloud, they face pressure to optimize the costs.
Main Cloud Costs
For the purposes of this article, we will approach this from the perspective of purchasing Infrastructure-as-a-Service level services, since that is the starting point for a complete package of services. Anything below that and you are either at SaaS, where you are using an application, or a basic service like DropBox for simple storage.
For most customers, running IaaS involves taking on-premises apps and moving them to the cloud for any of a variety of reasons, be it scale, flexibility, or to reduce costs. Because of this, the main costs for an IaaS subscription are compute and bandwidth. Compute costs will consume 60% or more of your total subscription. The next largest portion, bandwidth, can be as high as 20% depending on how much data you are moving between your on-premises systems and your cloud provider.
In both cases, there will eventually come a point where it’s no longer economical to operate in the cloud because the monthly costs exceed what you would pay to run it on-premises. In fact, Gartner says cost reduction should not be your motivation.
“We never advise customers to move to the cloud to save costs,” said Sid Nag, research vice president in the technology and service providers division of Gartner. “You don’t flip a switch and move to the cloud. You have to maintain old app on premises and are always running in a bimodal mode. Maybe four or five years later you start to see cost savings. In the first six months, absolutely not.”
Then there’s a hidden cost most firms don’t realize before moving to the cloud. Often, you can’t take an on-premises app and move it to AWS and expect it to run unchanged. At best, it won’t operate the same. Or, your app simply doesn’t take advantage of the characteristics of the cloud.
“Rewriting apps is a cost people overlook,” said Nag. “If you really want to take advantage of true cloud characteristics, you will end up writing an app that looks nothing like the app on-prem. It does the same thing but looks completely different. That costs.”
After compute and bandwidth, support can add up quickly, as can storage. Cloud providers allow customers to choose between flash and traditional disk storage, and the flash storage costs twice as much. Another cost that can spiral out of control is memory. Memory tends to be priced at a rate of 1:1, meaning if you double the amount of memory you allocate for your VMs, you will pay twice as much.
There are also a number of costs unrelated to the actual cost of using the service. They are:
- Over-provisioning: You allocate too much compute, memory and storage and pay for something you don’t use..
- Under-provisioning: The opposite problem, where you don’t have enough resources and either have to buy more or spend more to get the work done..
- Fire and forget: Spinning up a bunch of virtual machines and then forgetting to shut them down when you are done. If they keep running, the meter keeps ticking..
- Bad storage choices: You might get too much or too little or use the wrong kind of storage, since there are many to choose from..
- Non-cloud services: virtual hardware, like load balancers and VPNs, are often pushed on customers who think they need them and add up quickly.
- Exit fees: Say you pick a provider and decide you don’t like them and decide to move to another provider. You might get hit with a big fee to retrieve your data. Check for this in the fine print before signing on with a host..
- Support costs: Tracking down support issues can be complicated and your provider is going to charge you a hefty fee for it.
In this early era of cloud computing, many costs are higher than they need to be.
Cloud Cost Comparison
There are almost two dozen enterprise cloud services providers of note, so time does not permit going into a complete comparison of everyone. What we did was compare the top four – Amazon, Microsoft, Google, and IBM – along with two small players that may not necessarily on the radar of most companies but are extremely price competitive.
To start, we looked at the size of a Linux-based virtual machine in cores and memory. They range in size from small to double extra large and break down this way:
The price difference between the small players and big players, however, is astounding.
Is Internap the first cloud services provider that comes to mind? Or the second? But it’s clearly the leader in price. The provider 1&1, which primarily sells hosting and domains, is also extremely price competitive if your needs fit their system. That’s why it pays to shop around.
Next we look at Block Storage, a common form of cloud data storage. The prices vary widely, but also so do the services. Some providers offer much greater guarantee of performance, but it comes at a cost.
|1&1||Default SAN Storage||$0.07|
|Amazon||EBS Cold HDD||$0.03|
|EBS General Purpose SSD||$0.10|
|Standard Provisioned Space||$0.04|
|SSD Provisioned Space||$0.17|
|IBM||Endurance Block Storage||$0.15/$0.35|
|Internap||Persistant Block Storage||$0.30|
|Microsoft||Standard Disk LRS||$0.05|
|Standard Disk GRS||$0.10|
|Standard Disk RA GRS||$0.012|
There are many other price metrics, some easier to compare than others because the providers either do or do not providing pricing information on their Web sites. Data transfer rates, for instance, can be difficult to find for all the different providers because some providers don’t list them.
What these charts show is pricing can vary greatly, and the biggest providers aren’t necessarily the cheapest. So when researching cloud costs, it really pays to shop around and cast a wide net.