CURATEcamp 2010: Replication Breakout 8.16.10

Is it peer-to-peer or master-slave relationship?

Mark Phillips: microservices approach at UNTexas. Assumptions: boundaries of an object

Arsenic and a Cron job to replicate files doesn't necessarily work
because of physical drive arrangements between two sets of servers.
Instead they use a manifest and walk it to move the files.

Layers of
abstraction help with replication between a server and tape, etc.

After replication you want/need to do fixity checks to insure data hasn't been corrupted.

UCSD studied cloud costs for the super computing center, and found the costs were comparable to Amazon cloud services.

Fixity first, then replication, then fixity on the copy?