Comparing the cost of ALL competing cloud IaaS providers to find the cheapest… most of us would agree that that is a lot of work. But surely doing a reasonable comparison of just two cannot be much effort? Think again.
Let me enumerate just some of the challenges, for they are many…
A few of the challenges of cloud comparison
- Matching instance sizes. Some clouds sell fixed combinations of CPU, RAM and HDD, others offer you choice. Temptation is to choose cloud resources from the cloud that offers flexibility to match the combination imposed by the cloud that offers only bundled resources. This unfairly skews the comparison in favour of the likes of Amazon, where you effectively overbuy two of your three cloud resources.
- Matching SLA. If one cloud offers a better SLA than another, how do you price that benefit?
- Matching Price. Well how hard can this be? Guess what, this is actually quite complicated! Some clouds don’t include all the costs in the instance usage prices. Some costs that are incurred during usage of the instances may be split out. Many will probably be negligible for common use cases, but as a user new to cloud, how are you to know that when planning? I’m talking about the less than obvious stuff such as number of Gets, Puts and Reads. Amazon is an example of a cloud that breaks out many ancillary costs, presumably to avoid attracting all the bizarre use cases that can take unfair advantage of more simplistic pricing, and it isn’t their fault that people doing pricing comparisons sometimes miss this cost.
- Matching Payment Term. Whether you have to pay upfront or in arrears does matter. Your company has “an implied cost of capital”, i.e. the amount it would cost your company to borrow money to cover poorly negotiated payment terms. Prepayment vs 30 days in arrears for a startup using equity instead of debt might have a 30% cost of capital, so the cost is 60/365*30%=5%.
- Matching pricing period to load profile. If your workload spikes for say less than 10 minutes, and this is then not repeated during the same hour, there is nothing more important to your pricing than the time increment on the cloud’s billing. Take a cloud that works on calendar hours as the extreme example. Your workload runs for under ten mins over the hour…that will cost two hours or 120 mins. On a cloud such as CloudSigma, that would be just 10 mins due to their 5 minute billing cycle. So that is 1/12th the cost or a 91% saving.
I am quite sure that the five issues above are not the only five, but I think I’ve made my point that anyone who asks you to “just” compare the pricing on a few different clouds, clearly doesn’t understand the challenge that he or she is asking you to overcome!
Whether you are looking for the Best Cloud or the Cheapest Cloud for your use case, doing a cloud comparison is clearly not a simple task.
In a world of commoditising cloud infrastructure, it would certainly be easier if each user, to justify their choice of cloud, only had to compare their choice against a single standard cloud, and not against everything…