AWS
The reliability requirements are the same on every cloud. This page carries the AWS answers.
Verified against the platform, not against a deployment
DynamoAI runs on EKS and the requirements below are the same ones the platform needs anywhere. The commands are current AWS CLI forms. Where a row is blank, the equivalent has not been exercised on a real deployment and is not asserted here.
Verifying A Region
# Availability zones in the region
aws ec2 describe-availability-zones --region <region> \
--query "AvailabilityZones[].ZoneName" --output text
# GPU instance types actually offered, per zone
aws ec2 describe-instance-type-offerings --region <region> \
--location-type availability-zone \
--filters Name=instance-type,Values=g5.*,g6.*,p4d.*,p5.* \
--query "InstanceTypeOfferings[].[InstanceType,Location]" --output table
# Quota. List first, then read the one that matches the family you need.
aws service-quotas list-service-quotas --service-code ec2 --region <region> \
--query "Quotas[?contains(QuotaName,'On-Demand G')].[QuotaName,Value]" --output table
Tiers That Carry High Availability
| Requirement | AWS service | What satisfies it |
|---|---|---|
| PostgreSQL with high availability | RDS for PostgreSQL, or Aurora PostgreSQL | Multi-AZ. Aurora places replicas in separate zones by default |
| Cross-region PostgreSQL standby | Same | A cross-region read replica, or an Aurora global database |
| Redis with high availability | ElastiCache | Cluster mode with automatic failover, replicas in separate zones |
| Zone-redundant block storage | EBS | EBS volumes are zonal, and pin any pod that mounts them to one zone. Use EFS where a workload must move between zones |
| Zone-redundant object storage | S3 | Regional buckets are multi-zone by default |
Settings That Cannot Be Changed Later
| Component | Setting | Consequence |
|---|---|---|
| Node groups | Subnets, and therefore zones | Fixed at creation. Spanning more zones means a new node group |
| RDS | Multi-AZ | Can be enabled on a running instance, unlike some other clouds. Verify before assuming a rebuild is needed |
Region Pairs
AWS has no designated region-pair concept. Choose the second region on availability zones, instance availability and data residency. Cross-region replication is configured per service.