Synchronous TrueCopy ensures that an application write is not committed until the data has been written to both local and remote array. Asynchronous TrueCopy commits the transaction to the application when it is written to the local array. At some point following this, the local array sends the transaction to the remote array. If there is a link failure or disaster when asynchronous TrueCopy is used, it is likely that a small amount of data loss will occur due to the time lag between writes to the local and remote arrays.
If you are connecting to USP-V -- or Hitachi's new Virtual Storage Platform (VSP) -- external storage, you will have the option to use TrueCopy synchronous or asynchronous, but will also have the option of using Hitachi Universal Replicator (HUR).
HUR is also an asynchronous replication technology, although it uses journal volumes instead of cache to store the uncommitted transactions. The remote array, instead of the local array, also initiates data replication, which cuts down on the processor cycles required for replication on the local array. As always the decision on whether to use synchronous or asynchronous replication will rest on the application type and the distance between the local and remote locations.
The question as to which replication method is best for AIX servers is a difficult question to answer, as there really isn't a right or wrong method. What needs to be taken into context is the application profile and processor utilisation on the AIX host. The distance between the local and remote data centres may govern the question regarding synchronous vs asynchronous replication anyway, but if it doesn't, the application will be the deciding factor.
If the application cannot tolerate any missed transactions, synchronous replication will be the way to go, as this is the only replication technology that can guarantee that each transaction will be written to both arrays. The journal-based asynchronous replication of HUR can guarantee the transaction will get to the remote array only if the outage can be recovered from at some point.
It should be noted that synchronous replication will also guarantee to transfer any data corruption to the remote array, as it guarantees a copy of the local device. If corruption mitigation is the main use for the replication, asynchronous will be of greater benefit.
This was first published in October 2010