Metro Mirror Concepts

Metro Mirror can be run between two storage subsystems like DS8000s, or it can be run between two SVC SAN controllers. When run from SVCs, Metro Mirror is always Open Systems, but subsystem based Metro Mirror can be either Mainframe or Open Systems. Metro Mirror is often combined with Global Mirror to provide a 3 site failover system as shown in the diagram below. Here the primary and secondary sites are relatively close to each other and the data is Metro mirrored for zero data loss. There is a third, remote site which can be a considerable distance away and the data is cascade mirrored from the secondary disks to the 'tertiary' or R3 disks. This is Global or asynchronous mirroring to prevent an unacceptable performance overhead.

Metro Mirror, or Peer to Peer Remote Copy (PPRC) as it used to be called, is used to provide synchronous data mirroring between two different storage subsystems. IBM manuals often use Metro Mirror in the headings then PPRC in the text, which can be confusing. PPRC is a real time hardware function, the data is always kept exactly synchronised between the two boxes, and is independent of any operating system. The storage systems have to be connected by Fibre Channel and the boxes can be up to 300km apart without needing any extra hardware.

Reads are not affected by PPRC, they always come from the primary system. For every WRITE I/O, data is written to cache on the primary system, then replicated to the cache on secondary system. Without PPRC, the data write is complete once the data is stored in cache on the primary subsystem. With PPRC, a write is not complete until the secondary reports that the data is safely stored in cache. This can significantly add to the time taken for a write operation.

PPRC works at logical disk level, i.e. disks as they are defined to the operating system. The whole disk is mirrored. In a disaster, it will be possible to switch to the other set of disks and boot a system from them, without having to restore any data.
When PPRC is first initiated, all the data is copied. This can several hours if a lot of disks have to be synchronised

The following steps are required to set up PPRC

1. Lots of planning. I'd highly recommend that you configure your primary and secondary (and third if you have it) disk systems identically, and then mirror equivalent volumes. In any case, you will need to map these out.
   You will also need to work out your cabling requirements for connecting the subsystems together. Each physical path can host up to 256 logical paths. A physical path is bi-directional, but logical paths are uni-directional. It is possible to define paths from one logical subsystem (LSS) to up to 16 secondary LSSs, each of which can be in a different DS8000. However for simplicity its best to define each LSS with paths to its equivalent LSS in the secondary subsystem. One approach is to provide 8 physical channels or cables between a pair of boxes, then define 8 logical paths between each primary LSS and its corresponding secondary LSS, so that each logical path on a single LSS is on a different physical path, but physical paths can be shared between different LSSs on the same subsystem.
2. Now install and configure your hardware based on the plan you worked out above
3. Once you have all the physical hardware installed, you need to use PPRC commands to establish the logical communications paths between the storage devices. These commands are detailed on the next page. For failover purposes you will need to define two sets of logical paths, one running from primary to secondary, and the other from secondary to primary. This is because logical PPRC paths are unidirectional. In other words, they operate in only one direction. You can define logical paths in both directions on the same Fibre Channel physical link. You can establish up to eight logical paths between a primary LSS and a secondary LSS and then PPRC will balance the workload between them.
4. Once the paths are defined you need to establish the mirrored pairs. The PPRC commands to do this are detailed on the next page.

The following table explains some Metro Mirror terms and disk states

Primary disk	The disk which is addressed directly by the operating system. It will usually be in your main computer centre
Secondary disk	The mirror of your primary disk
Simplex	This disk is not part of any PPRC pair
Duplex	This Disk is part of an established PPRC Pair, and all is working ok.
Pending	A PPRC pair command has been issued, but all the data has not all been copied over yet
Suspend	This disk is part of a PPRC pair, but mirroring is suspended. The Primary disk will keep a record of updated tracks in a log, the data on the secondary disk will probably be out of step with the primary

If an error condition occurs and a mirrored volume cannot be updated, the primary disk will either go into an "extended long busy" or "queue full" condition and will pass an error message back to the host. This message can be trapped by automation to freeze all the disk subsystems.

back to top

---