|
|
|
|
|
|
by Animats
2915 days ago
|
|
|
That was written when Tandem was the leader in database replication. They had shared-nothing redundancy working very well. The problem was that their custom CPUs cost too much. Today, we see a lot of shared-nothing systems, where database slaves are kept in sync over the network. That's basically how Google stores data at scale. When you get big enough, you almost have to go that way. |
|
|