SyncMirror
SyncMirror can function in two modes:
- Local SyncMirror.
- MetroCluster SyncMirror (MC-Configuration SyncMirror).
And the difference between SyncMirror MCC and SyncMirror is that in the case of SyncMirror, there is constant mirroring of information from the NVRAM of one controller to two copies of the data in the partner controller of the HA pair. This approach is often used to protect the system from losing an entire shelf due to some failure.
In the case of SyncMirror MCC, NVRAM mirroring is performed simultaneously between several controllers: within the pair and on the responding site controllers. This solution is implemented using special FC-VI ports.
Features
. SyncMirror can perform replication virtually at the RAID level, and the most appropriate analogy that can be drawn here is RAID-60:
- there are two plexes (“sub-mirrors”), Plex0 and Plex1;
- each Plex can have several RAID-DP groups.
This analogy is valid because the pair of Plexes described above are defined as composite, mirrored parts of the same system. But in reality, in a stably running system, mirroring occurs at the controller, NVRAM level.
The key detail of SyncMirror is the necessity of full-dimensional symmetry of drives in a pair of journaled pools. Symmetry should be expressed in the same type, speed, volume, RAID groups.
Replication synchronization solves 2 problems:
- On the one hand, it allows to relieve the load from the disk subsystem, because it replicates only the memory itself.
- And on the other hand, to get rid of the problems of consistency and Split-Brain, you need to be sure that the information was written to the remote system.
The bottom line is that this synchronization, regardless of the type of storage system, can reduce response time and the time it takes to send information to the remote resource.
Maximum Seamless Management
SyncMirror data copies are managed using the same guidelines as any other data management in storage. In addition to ensuring high availability of information with SyncMirror, you can break the link between synchronized pairs of dataset copies with a single command so that those copies can be used by other applications in read-write mode.
This solution is almost ideal for avoiding the typical backup problems. Because background processes can work using mirrored information without interfering with the structure of the workspace. The result is a significant performance improvement.