Google announced their new Compute Engine pricing. While virtually everybody is focusing on significant price reductions, more important is their new model of sustained use discounts. Seems innocent but this might be complete game changer in the market.
Up until now, virtually everyone who planned to host application in any cloud had to make very basic decision – should I commit for long-term usage or go with on-demand servers. This could be reserved instances in case of AWS or longer period subscriptions in case of other providers. In either case, you had to think how you expect your future usage would look like. Therefore, while you had simplicity of consumption-based billing for on-demand prices, you still had to plan upfront to achieve greater cost optimization on expense of risk that predictions were wrong causing additional costs. This in turn was spoiling the perfect image of pay-what-you-use premise.
The first change came from AWS when it introduced Reserved Instance Marketplace over a year ago. It enabled users to sell their reservations in case predictions turned out to be invalid. This definitely improved the situation but still the risk of finding buyer was on the client, letting alone the 12% fee and sadly limitation to US customers only. Later there were further improvements of moving reservations between availability zones as well as changing instance types of a reservation within the same family. It got even better but still not yet where we all wanted to be.
Google is the first one to make it no-brainer for customers but still encourage predictable workloads. You just use on-demand servers. However, as soon as your server is run 25% of the time, a discount is applied. Then further with 50% and 75%. So the closer you get to the permanent usage, the cheaper it gets. Without any planning. Just based on actual usage. Like a self-cost-optimization plan. This is Google that takes the risk of usage predictability, not you.
This is not so noticed but amazing step of even easier usage of cloud services, bringing cloud elasticity to completely new level. I’m extremely excited to see what will be the response of other cloud services. I hope it is going to settle new standard that others would follow, making things even simpler.
Up until now, virtually everyone who planned to host application in any cloud had to make very basic decision – should I commit for long-term usage or go with on-demand servers. This could be reserved instances in case of AWS or longer period subscriptions in case of other providers. In either case, you had to think how you expect your future usage would look like. Therefore, while you had simplicity of consumption-based billing for on-demand prices, you still had to plan upfront to achieve greater cost optimization on expense of risk that predictions were wrong causing additional costs. This in turn was spoiling the perfect image of pay-what-you-use premise.
The first change came from AWS when it introduced Reserved Instance Marketplace over a year ago. It enabled users to sell their reservations in case predictions turned out to be invalid. This definitely improved the situation but still the risk of finding buyer was on the client, letting alone the 12% fee and sadly limitation to US customers only. Later there were further improvements of moving reservations between availability zones as well as changing instance types of a reservation within the same family. It got even better but still not yet where we all wanted to be.
Google is the first one to make it no-brainer for customers but still encourage predictable workloads. You just use on-demand servers. However, as soon as your server is run 25% of the time, a discount is applied. Then further with 50% and 75%. So the closer you get to the permanent usage, the cheaper it gets. Without any planning. Just based on actual usage. Like a self-cost-optimization plan. This is Google that takes the risk of usage predictability, not you.
