Similar to EC2 Availability Zones, it would be great for providing high availability if Linode was able to procure space in two, relatively close datacenters with inexpensive WAN links between them.

For example, HE has two interconnected Fremont datacenters, and could probably be convinced to charge a very low tariff for traffic between the two facilities. Latency would be low enough and throughput would be high enough that replication between the two datacenters could be done synchronously without issues. IPs would have to float across the two Fremont datacenters (I know this is not trivial, but should be much easier if the same carrier is used) so that load balancer failover would be possible.

This would allow a full site-wide outage to occur at either of the Fremont datacenters and with a proper HA setup, recovery should be quick. It's a solid guarantee that there is no sharing of any resources (rack, cooling, power, etc) between failover pairs.

+1

_________________
Got Escher? | @artagesw