Synchronous SnapMirror is a SnapMirror feature in which the data on one filer is replicated on another filer at, or near, the same time it is written to the first filer. Synchronous SnapMirror synchronously replicates data between single filers or clustered filers situated at remote sites using either an IP or a Fibre Channel connection.
Before Data ONTAP saves data to disk, it collects written data in NVRAM. Then, at a point in time called a consistency point, it sends the data to disk. When the Synchronous SnapMirror feature is enabled, the source filer forwards data to the destination filer as it is written in NVRAM. Then, at the consistency point, the source filer sends its data to disk and tells the destination filer to also send its data to disk. Finally, the source filer waits for the destination filer to acknowledge that it sent data to disk before continuing with the next write.
If there are problems with your network, your synchronous replication might go into an asynchronous mode. Ordinarily, the source and destination filers periodically communicate with each other. In the event of a network outage, synchronous SnapMirror goes into an asynchronous mode if the periodic communication is disrupted. When in asynchronous mode, the source filer tries to communicate with the destination filer once every minute until communication is reestablished. Once reestablished, the source filer asynchronously replicates data to the destination every minute until a synchronous replication can be reestablished.
You should consider the following criteria before using the Synchronous SnapMirror feature:
Synchronous SnapMirror is not supported on the following filers:
You decide which data you want synchronously replicated. For example, you might synchronously replicate database data and asynchronously replicate home directories. If the home directories contain important log data, you might synchronously replicate it, but adjust the synchonicity of the transfers to maintain performance levels. Synchronous SnapMirror allows a user-defined lag time of a number of operations or milliseconds before transferring data. See Controlling the synchronicity level to improve performance for details.
Ideally, the source filer and destination filer should have the same configuration (type of filer and disk geometry).
The type of filer and disk geometry of the destination filer impacts the perceived performance of the source filer. Therefore, the destination filer should have the bandwidth for the increased traffic and for message logging. Log files are kept on the root volume; therefore, you should ensure that the root volume spans enough disks to handle the increased traffic. Network Appliance recommends that your root volume spans four to six disks.
Synchronous SnapMirror can support traffic over Fibre Channel and IP transports. SnapMirror also allows multipathing giving you the ability to either balance the load between two paths or reserving the second path for failover. For optimizing performance, you can use the best route available between the source and destination filer and you can restrict the route to traffic between the two filers. See Setting up Synchronous SnapMirror using multiple paths for details.
The following sections discuss how you can create and modify synchronous SnapMirror relationships:
Copyright © 1994-2003, Network Appliance, Inc. All rights
reserved.
Part No. 210-00476
File: mirror26.htm
Last updated: 11/13/03 17:17:26