Fixed or Variable?

No, this is not about taking out a new mortgage. In the world of cloud computing, a management shift is taking hold. At Snowflake, we provide customers with a cloud data platform we charge for based on actual usage. You can consume and be invoiced for as much, or as little, as you use. The model is granular, so you’re charged by the machine second. And because Snowflake is highly elastic, you can run massive amounts of computational capacity for short periods of time. 

This is more equitable than the subscription-based SaaS model in which customers are charged a license fee for a specific term, regardless of how much they actually consume of the service. Like minutes on your mobile phone: you don’t incur cost until you start consuming that time.

 In IT, we’ve spent a lifetime optimizing workloads in a fixed capacity environment. Adding to that capacity did happen but rarely, because it was difficult and costly. So, most of the time, we worked with fixed limits on storage and compute. We became experts at tuning and tweaking, squeezing every ounce of workload out of our capacity. Efficient and powerful operating systems and databases were instrumented for system and database administrators, who wanted the best performance. 

 Enter the public cloud. Capacity is no longer fixed. It’s now variable. You can consume as much storage and compute as you are capable of invoking (and hopefully paying for). Initially, this newfound freedom didn’t fully register. The fixed capacity mentality lingered in the cloud because on-premises systems, now hosted in the cloud, were not designed for variable consumption of compute and storage. 

 Enter Snowflake. A cloud data platform with no on-premise or fixed capacity heritage. It was literally re-imagined for a cloud scale environment where storage and compute are variable, virtually unlimited, and invokable on an as-needed basis. For as long or as short a duration as the workload calls for. Check, that all sounds great, where do I sign up?

 Just as in mortgage origination, variable rate loans are great as long as rates are going down. In the world of cloud, everything is great, paying only for what you use, but once the capacity lid is removed, users discover they can now fire up new workloads at will, provision them with ever larger amounts of compute power, and have a field day running wild in the unlimited world of cloud. 

 And then the bill comes. Database and systems people understand this. CFOs, not so much. They only look at what you spent in the current period as compared to the previous one. What could possibly be the explanation for this? And so a whole new management challenge is emerging: how to manage variable capacity?

 New features have since arrived to provide user chargebacks, hard and soft governors on consumption, alerts and notifications, and consumption dashboards. All of this is helping, but genie is out of the bottle, and users throughout the enterprise need to govern their consumption in ways they have never had to before. You simply ran out before. Now, you don’t.

Variable capacity begs new questions, such as what are appropriate levels of consumption? Initially, we see customers going to town on their backlogs. But then the potential for further transformation kicks in. We are spending more in relative terms. Not because we can but because the business deems it necessary, if not vital. On an apples-to-apples comparison, cloud computing is much cheaper than on-premise environments because you only pay for what you use. Let’s face it, we only had to manage the capacity our own data centers allowed. We neither had the scale nor the where-for-all for it.

 CFOs, while not amused, have to make more nuanced and thoughtful judgements on how to think about consuming capacity in the cloud. Things are now possible at a scale and speed that were unimaginable before. The potential is limitless. We can’t just open the flood gates but we do have to pick battles worth fighting. Because the competition likely will.