...
- Even though the current copy of information is in main memory, data must be just as durable as if it were stored in a traditional disk-based database.
- The durability of data must be guaranteed before a write request returns.
- With thousands of servers in the cluster, machines will die constantly; it's possible that the system will almost always be in a state of crash recovery.
- Recovery from "simple" failures must be instantaneous and transparent.
- Data consists of small chunks of information (a few hundred bytes), updated in their entirety.
Failure scenarios
We might want RAMCloud data to survive any or all of the following scenarios:
...