How to apply cost saving patterns for Atlassian Data Center on AWS deployments

Goal

You want to facilitate the Atlassian Data Center on AWS quick starts for performance and functional testing without breaking the bank.

AWS (and other cloud providers) offer significant cost saving options, but this usually requires some proficiency
Atlassian does not seem to be overly concerned with paying more for AWS usage than necessary
- likely remedied by large scale user discounts and purchasing options (reserved instances)
- also applies to JPT I think
Diverging quick start branches complicate collaboration
- this will hopefully get resolved down the road

	Cost \| Features	Landing page	Repositories
AWS	\|	https://aws.amazon.com/quickstart/	https://github.com/aws-quickstart?q=atlassian
Atlassian	\|	Atlassian Data Center on AWS quick starts	https://bitbucket.org/atlassian/atlassian-aws-deployment/
Utoolity	\|	TBD (community summary)	https://bitbucket.org/utoolitylabs/profile/repositories?search=quickstart

Option / Status	AWS (official)	Atlassian (upstream)	Utoolity (fork)	Notes
Resource configuration
Facilitate current instance types				those are usually more cost-efficient
Default to inexpensive parameters				many tests do not require production oriented environments (e.g. multi-az)
Facilitate EC2 spot instances ('use spot')			\|	significant savings (~75%) meanwhile easy to use See https://www.infoq.com/news/2018/01/amazon-ec2-spot-streamlining BBS NFS instance pending (non-trivial) single EC2 instance is not HA!
Resource utilization
Reuse/Disable optional components				e.g. NAT gateways and bastion hosts
Stop cluster during off-hours ('cold standby')			\|	significant savings (~75%) RDS instances pending How to mitigate the automatic weekly RDS instance restarts BBS NFS instance pending
Terminate cluster during off-hours (fully automated backup/restore)				savings potential depends on MTTR goals (non-trivial)
Reserved instances	N/A	N/A	N/A	for savvy AWS users only