IBM DS6000 Manual Do Utilizador

IBM System Storage DS6000 Series: Copy Services with IBM System z

When replicating data over long distances, beyond the 300 km, asynchronous data 
replication is the valid approach. This is basically so because with the asynchronous 
techniques, the application I/O processing at the local storage disk subsystem (storage 
server) remains independent of the process of data transmission to the remote storage disk 
subsystem. 

Still with the asynchronous data replication techniques, we need to provide additional means 
to ensure data consistency at the remote location. It even requires a solution that guarantees 
data consistency not only within a single local-remote pair of storage disk subsystems but 
also across multiple local and remote storage disk subsystems. 

For a given pair of local and remote storage disk subsystems, a time stamp approach leads to 
consistent data at the remote storage disk subsystem. Using this approach, and by sorting 
the I/Os by their time stamps, the write I/Os can be applied at the remote disk subsystem in 
the same sequence as they arrived at the local disk subsystem. 

But when the application volumes are spread across multiple storage disk subsystems, this 
time stamp concept alone is not sufficient to replicate data and provide data consistency at 
the remote site. This additionally requires a Consistency Group concept. z/OS Global Mirror, 
formerly known as XRC, is an example of such a solution and utilizes the concept of 
Consistency Groups combined with time stamped write I/Os for all involved primary volumes 
within a Consistency Group or session. See Part 6, “Global Mirror” on page 265 for more 
information regarding z/OS Global Mirror.

In the rest of the present section we discuss how consistent data, that is dependent writes, 
are managed either with a synchronous technique like Metro Mirror versus different 
asynchronous techniques like Global Copy or Global Mirror.

In normal operations, the nature of synchronous data replication preserves data consistency 
for 

dependent writes and data consistency are explained in detail in 

Secondary

Primary