Availability and Recovery: Discussion

Today was a rare "purse discussion". We discussed strategies for maintaining availability in light of failure. We discuessed the role of specialized hardware (load-balancing switches) directory services (DNS), repliction, consistency models (allowing some staleness, when possible), coordination (elected and appointed coordinators, truly distributed decisions), understanding natural partitionings of data and services, and humans in maintaining system availability.

If you missed it, it wasn't really test material, per se. It was mostly a chance to step back and see how the piecves go together. Ask a friend. Or, ask yourself, how does Amazon work? If you are feeling guilty, check out some of what they've written on the subject, by the way. They are in the business of selling the model to others, as it turns out :-)