Synch v. asynch replication
Replicating data on a separate storage device is a popular way of ensuring data availability. While replication doesn't replace backup, switching to the replicated data in the event of a device failure is faster than restoring from backup. But when do you write the data to the replicating device? That depends on a number of factors, including distance and bandwidth.
Replicating data on a separate storage device is a popular way of ensuring availability of data. While replication doesn't replace backup, switching to the replicated data in the event of a device failure is faster than restoring from backup. But when do you write the data to the replicating device? That depends on a number of factors, including distance and bandwidth.
The most obvious answer to the question is "immediately". That is that data is written to both devices synchronously and the write operation isn't regarded as complete until the data is recorded on both devices. This is the method most commonly used for mirroring when the devices are in the same system, or even part of the same RAID array. It is fast and simple (as such things go), however it assumes a reliable, high-bandwidth connection to both devices. Otherwise performance is seriously degraded, or the write even fails entirely.
If the data is being replicated at a remote site, asynchronous replication is often a better solution. In asynchronous replication the writes to the local and remote devices are separate operations and often the remote write isn't initiated until the local device reports its write is complete. In some cases the writes to the remote device are deferred until the local data has been confirmed as good. Asynchronous replication is especially useful when the connection between the local and remote sites has a relatively low bandwidth, 45 Mbps or less.
Asynchronous replication introduces a number of complexities in storage, including dealing with the possibility that the data is corrupted while buffered for transmission to the remote device, failures in the long-distance link and the need to keep logs and dataspaces synchronized for RDBMS operations.
Sun Microsystems gives a good explanation of synchronous and asynchronous replication in its white paper on its StorEdge Network Data Replicator Software at www.sun.com/storage/white-papers/sndr.html.
About the author: Rick Cook has been writing about mass storage since the days when the term meant an 80K floppy disk. The computers he learned on used ferrite cores and magnetic drums. For the last twenty years he has been a freelance writer specializing in storage and other computer issues.