The Business of IT

Cloud pricing wrinkles and tea in China

As you’ve probably heard, Amazon Web Services reduced their on-demand cloud prices significantly last week. You’d think customers would be happy across the board, but that’s not the case. Here’s why, and what will happen as a result.

As discussed previously, AWS customers translate their capacity planning into Reserved Instance purchases, based on the relative savings these RIs provide over on-demand prices. But, when on-demand prices are reduced without a corresponding reduction in the instance-hour price for RIs – as happened last week – the RI breakeven point shifts and upsets the optimal RI coverage calculus. AWS customers who purchased RIs before the price reduction can find themselves stuck with inventory that now costs more per hour, in amortized terms, than the cost if they had not purchased the RI. I have several clients in this situation, and none are very happy about it.

Tea in China

I can imagine the counterargument to this thinking, no doubt coming from the mouth of a cool-headed economist: If the customer was happy when she bought tea at a 10% discount, why should she be any less happy when the price is further reduced a day or a week later? She deemed the value of the tea to be worth the cost when she bought it, and that value has not changed. By rational reasoning, she should be equally satisfied today.

But people don’t think purely in economic terms – emotions play a part in their decisions as well. And the AWS customers who are feeling unloved by this price reduction are the very customers, I imagine, that AWS wants to keep most: These customers have made long-term commitments to AWS already. So, it won’t be long until we hear another price reduction announcement from AWS, specifically directed at this customer segment. Don’t be surprised to see AWS granting a proportional hourly price reduction – or an equivalent credit – for already-purchased reserved instances.

The Business of IT

Cloud price reductions and capacity planning

Last week both Google Cloud Platform and Amazon Web Services reduced their prices for cloud computing services significantly to comparable levels, and both now offer significant discounts for long-term usage. Yet, though the two cloud services may seem similar, their radically different long-term pricing models reveal just how different these cloud offerings really are.

Whose responsibility is capacity planning?

The core difference between GCP and AWS is in capacity planning: Whose responsibility is it? In AWS, the customer owns their own capacity planning. If the customer can accurately predict their needs for the long term, they can purchase Reserved Instances and save significantly as compared to the on-demand cost. Whereas in GCE, Google owns the capacity planning. GCE customers are granted a Sustained Use discount at the end of the month for resources that were active for a significant portion of the month. The GCE customer might track their expected vs. actual costs and be pleasantly surprised when their bill at the end of the month is lower than expected, but the GCE customer cannot a priori translate their capacity planning prowess into reduced costs.

This begs the question: Are cloud consumers actually good at capacity planning? Sadly, no. Capacity planning in the pre-cloud age of the data center was dicey at best, with managers relying on overprovisioning to save their necks in the face of so much uncertainty. This overprovisioning is no longer a time-to-market-driven necessity in the cloud – but classic capacity planning is no more accurate in the cloud than it was in the data center. That’s why a host of new cloud financial management tools, such as Cloudyn, have sprung up in recent years: These tools help the cloud consumer predict usage and optimize their up-front commitments to maximize cost savings.

Using cloud financial management tools can help you reduce long-term costs, but only to the extent that the cloud provider rewards accurate capacity planning with lower prices – as does AWS. With Google Cloud, there’s no such pricing lever to pull.

But, make no mistake: Choose your cloud provider based on technical and business merits, not based solely on the long-term pricing model. Contact me if you need help.