2009-08-23 21:25:46 UTC
I am looking into implementing a DFS Replication setup for a system I
administer. Depending on the time of the year, I would have as little
as 3 and as much as 9 servers needing to have about 600GB (currently,
will grow) of data replicated between them all. As it stands right
now, changes to the files can occur on any server so there wouldn't be
any authoritative node that would have all changes made on it. The
files needing replicated can be anywhere from 5KB to 7GB.
My question is what would be the most reliable and best performance
way to setting up the replication connections? Would it be best to go
with a full mesh topology where all all members replicate to each
other. Or would be better to have each member replication to two or
three other members, making sort of a circular topology but setup in a
way where no single member could would break the replication.