Availability
Tied to each cloud provider, AWS or GCP only
All public clouds, on prem and local
Architecture
Designed as single instance architecture
Architected to span globally distributed data centers, yet still valuable in a single data center
Scale
Scale only through bigger machines or operationally intense manual sharding
Elastic and auto scale means minimal-to-no operations
Resilience
Manual operations required for active-passive failover and complex remediation
Naturally resilient to outages with granular control at the row level
Required downtime
Database must be taken offline for upgrades and schema changes
No downtime upgrades and online schema changes
Transactional consistency
Configurable to guarantee consistency but only in a single instance
Guaranteed correct data in single data center and beyond
Multi-region
Limited ability to optimize latency concerns for multi-region writes
Advances in parallel commits and transaction pipelining reduce broad geographic transaction latencies
Data residency
Entire database lives in a single region; data residency is not an option
Data is tied to a node at the row level to optimize performance and/or regulatory compliance
Hybrid and multi-cloud deployment
Tied to explicit instances only, no options for multi and/or hybrid cloud
Run anywhere on any cloud, even across regions and clouds
Migrations
Simple and fairly straightforward
Wire compatible with PostgreSQL with some required work for syntax
Freedom
Tied to a public cloud provider
Run anywhere and across multiple clouds
Pricing
Simple straightforward pricing, with complexities around active-passive systems
Simple, straightforward pricing, plus the ability to tie data to a location to avoid egress costs