[FOLIO-2057] SPIKE: explore AWS K8s cost models Created: 31/May/19 Updated: 03/Jun/20 Resolved: 13/Jun/19 |
|
| Status: | Closed |
| Project: | FOLIO |
| Components: | None |
| Affects versions: | None |
| Fix versions: | None |
| Type: | Task | Priority: | P3 |
| Reporter: | Jakub Skoczen | Assignee: | Ian Hardy |
| Resolution: | Done | Votes: | 0 |
| Labels: | platform-backlog | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original estimate: | Not Specified | ||
| Issue links: |
|
||||||||||||||||||||||||||||||||
| Sprint: | CP: sprint 65 | ||||||||||||||||||||||||||||||||
| Story Points: | 3 | ||||||||||||||||||||||||||||||||
| Development Team: | Core: Platform | ||||||||||||||||||||||||||||||||
| Description |
|
Cost will vary depending on the approach:
Factors to take into account:
|
| Comments |
| Comment by Ian Hardy [ 05/Jun/19 ] |
|
Doing some initial exploration it looks like RDS instances: Using m4.large size instances (2 cpu, 5 gb ram, 500) for example, a naive estimate for monthly compute/storage costs might be, ec2: RDS: RDS pricing is broken up into Database storage, Backup Storage, and Data Transfer. Database storage: Backup Storage: An initial guess is that we could pay up to 2x the upfront cost of running ec2 instances with the same computer resources/storage (not factoring any additional backup storage for ec2) for the benefits of RDS. If our initial goal is to provide storage for the CI environment it seems like RDS is not cost prohibitive and offers some real savings on the labor required to sysadmin PSQL on EC2. |
| Comment by mark.stacy [ 10/Jun/19 ] |
|
Ian Hardy I have provided information on EKS and EC2 K8s costs within a markdown gist. https://gist.github.com/mbstacy/50e5cfa5c4faad560c513595977ad096 |
| Comment by Ian Hardy [ 11/Jun/19 ] |
|
Thanks for the writeup Mark. A few other updates since this ticket was first written. >number of clusters (potentially low for development purposes) Since we're running the one cluster, I think it would make sense to EKS (already using) and RDS. Should there be a need to run a large number of database instances in the future we could try to optimize for cost by using ec2. It seems reasonable to me to start with reserved reserved instances for worker nodes. There's some additional overhead to use spot instances to make sure the cluster can handle an instance disappearing on short notice |