On: IBM's Lost Decade

[wmf]

Speaking of k8s markup, AWS only charges for the masters and they recently cut those prices by 50%. This problem doesn't exist.

[ablekh → wmf]

Why do you think AWS charges for K8s master nodes (control plane), when both of their major competitors do not?

[wmf → ablekh]

Because they can, of course. But it's still hard to imagine a third-party product (e.g. OpenShift) being cheaper than $0.10 per hour.

[ablekh → wmf]

They surely can. But so is true for Azure and GCP as well. Thus, I'm curious about whether it is a feasible pricing strategy for AWS, considering the fact that EKS is not much different from their competition's managed K8s offerings.

[wmf → ablekh]

I mean AWS can charge more because they're viewed as the "default" or "leader" cloud; Google and Azure are not.

[ablekh → wmf]

Understood. This IMO is a bit of a shaky ground, because, while still being (or being perceived as) a cloud leader - though Azure a formidable competitor across the cloud spectrum - AWS is definitely not a leader in the managed K8s area (where GCP clearly is, at least, as of now).

[closeparen → wmf]

EC2 costs almost twice as much per core as my company’s in house datacenters, and that’s before bandwidth.

[adrianN → closeparen]

But can you just turn your datacenter off and not pay for it when you don't use it? Can you double the number of cores in your datacenter by running a script and waiting five minutes?

There is little point in using EC2 if you never want to do either, I agree. But it seems like many people do and are willing to pay more per core to do so.

[closeparen → adrianN]

The idea of a hybrid cloud solution is to pay EC2 rates for only the unpredictable portion of your workload.

It is in fact a little cheaper to keep 2x spare capacity on hand vs. run the 1x workload on AWS all the time.