Originally posted by: Chris M Evans
Here's the question; how does TrueCopy manage the replication of data between subsystems?
I suspect each write to the primary cache for a volume which is replicated is tagged with a timestamp. Each changed block of data is then pushed to the secondary system in sequence; once confirmed at the remote site, the I/O is signalled complete to the host. That's all straightforward, however...
How is the data batched to make best use of the TC links which are defined? As more writes are recieved, are the writes batched to send to the remote host? How is this affected by the differing block size writes of primary hosts? Does TC even bother to batch or does it simply regurgitate the writes out of the TC links to the remote host?
This is important to me as I'm attempting to understand the impact of long distances on TC replication; if TC does attempt to batch data, then potentially as a subsystem gets busy, the replication process would become more efficient over distance; imagine a very quiet system where writes occur infrequently; these would have to be simply reproduced as is over the TC links.
This all refers to USP/NSC if I haven't already mentioned.