Redpanda on EKS costlier than EC2?

[pdeva]

EKS currently doesn't have a CSI driver for local SSDs. Thus workloads on EKS are forced to use EBS disks.

This means that deploying Redpanda cluster on EKS, it is forced to use EBS resulting in:

1. additional network costs for data transfer to/from EBS
2. higher per gb cost of storage
3. higher latency than local ssd

The use of EBS doesnt buy any additional durability guarantees either since Redpanda will replicate data across azs anyway.

Am I correct that considering all of the above, it is better to deploy Redpanda on EC2 instances since it can make use of the local SSD of the instance, thus resulting in being both cheaper and faster than a similar installation on EKS?

[emaxerrno]

We use local disk on EKS on our cloud.

[emaxerrno]

Deployment automation depends on your comfort. The absolute easiest thing is apt-get install on 3 VMs really. Redpanda is quite stable in that way

but if you use k8s and are an expert in k8s, you can get it to work there. just a lot more sophistication on your end.

[pdeva]

could you share what solution/helm-chart you use for using local disk on EKS?