Hello, I would like to know how it is possible to maitain consistent replicated data, in a DHT replication system.
For example. Assume 3 nodes (A,B,C) that form a replica group for some key. A Put(key,val) is performed, through the master (say A) and is propagated to B and C. Assume C crashes before it gets the Put. His successor (in Chord), say D, becomes responsible of that key. The replication system will make D recover the key we talk about. But what if, the item related to the key recovered is of a previous version because the recovery took place at B before the Put arrived at B? I mean, D will never get the Put. So it will remain with an out-of-date data. Any algos for consistent atomic writes? Thank you. _______________________________________________ p2p-hackers mailing list [email protected] http://lists.zooko.com/mailman/listinfo/p2p-hackers
