Split operation technically possible (run in 2 DC’s)
Network: Single Point of Failure, sustainable IP-addressing scheme, Routing, Security, Bandwidth, Latency
Compute: Tightly coupled nodes, Infiniband, Latency, Shared Storage
Downtime
Reduced performance
Reduced redundancy
Timing
- Hardware written off, time to replace
Costs
- Additional costs/investments/temporary hardware
- Effort
Risk
Cabling