Yeah I wasn't there from day 1 but I think the team I'm on was never doing F#. Still, I don't know if being there from the beginning would be a huge feather in my cap given that the only person who came out looking good out of this thing was Mark Lore.
@HoXw11 lol. AWS has EKS, a "managed control plane". It is an absolutely useless service; the user is still responsible for managing the nodes, setting up the ASG rules, joining the nodes to the cluster (installing and configuring kubelet). I won't even get into the lack of a supporting ecosystem from AWS, such as terrible/utterly lacking IAM integration, container-or-even-pod level management and monitoring, etc.
You're fooling yourself and nobody else if you're going to argue that AWS has a comparable offering to GCP, or even Azure, which has both a younger and more complete entrant into the market of "managed Kubernetes".
@fahrenheit if one must to run containers on AWS (because they're already locked in and moving to k8s on GKE seems insurmountable), using ECS with a deployment type of EC2 or Fargate is the best option, although far more expensive, restrictive, and unnecessarily complex and more difficult for beginners to grok (although this is true of much of AWS and not just ECS)
It's short-sighted to rely on a service provider by your competitors, because of the amount of lock-in that occurs and the amount of specialized assistance such large companies would require in order to fully move to the cloud (it's never actually as seamless as advertised.) Amazon could very easily sabotage Walmart's data strategy if they were stupid enough to try to put their data on the Amazon cloud.
@tt10 I don't think AWS have significan economical advantage to offset the potential risk. it's short sighted to compare Netflix va amazon to retailers vs amazon. What's core business to each of them, who has upper hand in each field, if possible conspiracy happens on spying, what's the cost of it all have to be assed. In short, it's not worth giving amazon money which will be used by amazon to compete with the same company. Ya... Amazon will still make money regardless....but nothing wrong if retailers decide not to contribute to Amazons growth by all means.
I also wonder if Netflix would choose AWS, if they have to start re building their platform today. Content they had and what amazon had when Netflix onboarded AWS is not the same today. Regardless, retailers won't see in the eyes of Netflix.
AWS because it has almost every solution built-in. Most medium complexity to advanced use cases can be production-ized within a few days using easy to consume wrappers and self managing infra underneath. The cost of running your services is low and production support is by far the most superior.
However, if you do not see yourself using a whole lot of integrated cloud services in the future and do not mind managing some of your infrastructure by yourself, you can consider alternatives.
GCP, because I know Amazon's unbelievable amount of technical debt inside and soon enough reality will catch up. And their closed mindset and bad treatment of engineers can't have any future. Just waiting to see how long until they run out of steam.
If you plan to grow to the point at which Jeff Bezos gonna notice, think twice before using AWS. If something were to happen to your contract, migration to other cloud providers could drag you way back and cost you customer base.
It is not Kubernetes btw, people commenting on kubernetes do not get the true nature of depth in cloud services. AWS has multiples of custom services that will make a retailer’s life a heaven-why they have so many services for retailers? - must be one of their key retailer customers...(any guess?) Take dynamo dB and imagine a xmas season without ddb, or managed elasticache - this service will be 10x important when you are in prod (I know retailers switching just on this one since GCP doesn’t have this) and there are 100 more custom services from call centre services to mobile app testing services which will all save $$ millions for retailers like Macy’s - do not limit your cloud conversation on compute...Compute platforms are a sign of the past and the investment on them assuming that customers will consume the compute in terms of server or a container within the next 10 years is a strategic mistake.
What? The example layers of abstraction that I gave are zero-cost and work perfectly fine.
They aren't necessary, and anyone who wanted to could instead choose to use a service provider's service directly. An example of this might be RDS or Google Cloud Spanner -- there aren't equivalent abstractions yet.