Right, and a lot of the cloud naive make lofty positive assumptions that aren't true.
For example, you can't lose data in the cloud.
There are very much operational mistakes you can make with combinations of S3 settings around versioning and deletes, that result in permanent irretrievable dataloss.
My last shop managed to do this and then was implementing some sort of cloud data backup scheme.. in the cloud, lol.
The other assumption is that sure compute is costly, but since you can spin it up&down you'll save so much. As it turns out, most apps, most of the time, do not have bursty use cases that merit paying 2-3x for compute in hopes of spinning it down when idle. The funny thing is the same people selling this line are also the ones telling you to negotiate savings with some of those annual agreements that require a minimum amount of compute/spend.
The last one is assumptions about hockey stick compute growth needs, and that of course choosing AWS will make this easier than having to constantly procure servers. Maybe! But few have hockey stick compute growth needs for long. And it's not that impossible to trade your servers out every 18 months to get your compute density growth. And you aren't always guaranteed AWS compute at prices you want, as we've seen shortages of certain classes compute and needs to reserve up front, etc.
For example, you can't lose data in the cloud. There are very much operational mistakes you can make with combinations of S3 settings around versioning and deletes, that result in permanent irretrievable dataloss.
My last shop managed to do this and then was implementing some sort of cloud data backup scheme.. in the cloud, lol.
The other assumption is that sure compute is costly, but since you can spin it up&down you'll save so much. As it turns out, most apps, most of the time, do not have bursty use cases that merit paying 2-3x for compute in hopes of spinning it down when idle. The funny thing is the same people selling this line are also the ones telling you to negotiate savings with some of those annual agreements that require a minimum amount of compute/spend.
The last one is assumptions about hockey stick compute growth needs, and that of course choosing AWS will make this easier than having to constantly procure servers. Maybe! But few have hockey stick compute growth needs for long. And it's not that impossible to trade your servers out every 18 months to get your compute density growth. And you aren't always guaranteed AWS compute at prices you want, as we've seen shortages of certain classes compute and needs to reserve up front, etc.