We need to each employ individual good practices for maintaining hygiene of any shared accounts we are using
- If we don’t want things to get out of hand as the team grows, the general rule of thumb to follow is ongoing maintenence. This means:
- Cleaning up any unused resources (instances, buckets, volumes, key pairs, etc.)
- Tag your user handle to resources (EC2 instances, EBS volumes, snapshots, etc.). In particular, things like EBS volumes and snapshots tend to build up so it’s important to maintain these (i.e. delete any unused volume and older snapshots)
In AWS, use this to choose the region to operate in:
- Stick to N.Virginia (us-east-1) for cheapest compute:

-
For choosing instance types:
- For CPU intensive requirements, choose C type instances which are optimized for compute heavy workloads. Good instance types in this category in terms of price/performance include C4 and C5 instances
- For instances that only require burstable compute, choose T instances. T3 instances provide a less powerful, cost-effective alternative to the fixed M5 instance.
- For instances that require GPU compute, choose G type instances. G4 and G5 tend to give some of the better price/performance value in this category
-
We should be diligent to stop our instances overnight when not using them. To prevent dynamic allocations of IPs between restarts of an instance, consider associating Elastic IPs to instances to allocate a dedicated IP (0.5 cents/day)