Vlad takes us on an interesting measurement journey into alternatives for minimizing your public cloud computing costs while also maximizing your customers' delight with your low-latency service by auto-scaling your kubernetes services more effectively using Fargate.
He links to this talk by Claire Liguori on interactive workloads.
Among the (very-many) disadvantages of on-premise systems at old stodgy companies is that the sunk costs of hardware and data centers cannot easily be flexibly allocated and "charged back" to the consuming entities within the company. On the public cloud, if vendor-autoscaling causes some latency or over-provisioning for a few minutes, you can use a serverless solution like Fargate to reduce the waste and customer pain.
No comments:
Post a Comment