Metro Mirror, Global Copy, and Global Mirror Quick Reference Background Information Dependent write means that the start of one write operation is dependent upon the completion of a previous write to a volume in either the same disk system or a different disk system. They are the basis for providing consistent data. Consistency means that the order of dependent writes is maintained. With consistent data at the remote site, it is possible to do a data base restart, which is much faster than data base recovery. Asynchronous processing means the separation of data transmission from the signaling of I/O complete. The order of dependent writes is not maintained, so data is inconsistent. Manual intervention is required to create consistent data. Distance between local and remote has little impact upon the response time of the primary volume. Synchronous processing is when the I/O complete is not returned until the update is transmitted to the remote site and acknowledged by the remote. Recovery Time Objective (RTO) The elapsed time after an outage before the Application is back up and running. (Server/Network/Workload & Data all available again.) Recovery Point Objective (RPO) The amount of data lost in the event of a disaster. (Transactions that had been completed but we lost as they have not been reflected at the recovery site.) PPRC Rules PPRC secondary volumes must be off-line to all attached systems One-to-one PPRC volume relationship Only IBM (DS8000/DS6000/ESS)-to-IBM supported PPRC ports defined as SCSI-FCP Logical paths must be established between LSSs FCP links are bidirectional and can be shared with server data transfer Up to 8 links per LSS More than 8 links per physical subsystem is allowed Up to 4 LSS secondaries can be connected to a primary LSS A secondary LSS can be connected to as many primary LSS systems as links are available Distance FCP links 300 KM without RPQ Metro Mirror Uses synchronous processing to mirror data The sites using Metro Mirror must conform to the following conditions: Can accept some performance impact to application write I/O RPO=0 possible using automation Distance within maximum limits 300 KM fcp links without RPQ Metro Mirror provides consistency by: Specifying consistency group on the establish path command If an error occurs between local and remote sites, the pair will suspend with messages or SNMP alerts being issued. Automation catches the alert and issues FREEZE/RUN commands to LSSs that have dependent data. The FREEZE command also deletes the paths from the LSS. Writes to these volumes will receive extended long busy (CKD) or I/O queue full (Open) and dependent writes will be queued. Once all of the LSSs have been frozen, you can issue RUN commands to them to once again allow writes. Otherwise the condition lasts 2 minutes. This solution requires automation when consistency is desired across multiple disk systems and is highly recommended in one-to-one environments. Planned outage for testing: FREEZE/RUN to LSSs to suspend pairs (also deletes paths) Failover to B volumes FlashCopy B to C volumes Establish paths from local to remote Failback A to B volumes Test on C volume Metro Mirror, Global Copy, and Global Mirror Quick Reference This document covers: Dependent writes and consistency PPRC rules Metro Mirror Global Copy Global Mirror How each solution maintains consistency Planned and unplanned procedures for each solution March 2007 © International Business Machines Corporation, 2007. IBM, and FlashCopy are registered trademarks of International Business Machines in the U.S. and other countries. Other company, product, or service names may be trademarks or service marks of others. IBM reserves the right to change specifications or other product information without notice. This publication may include typographic errors and technical inaccuracies. The content is provided as is, without express or implied warranties of any kind, including the implied warranties of merchantability or fitness for a particular purpose.

Global Mirror for zSeries (XRC) Metro Mirror Unplanned outage: Automation detects problem and issues FREEZE/RUN commands to LSSs and pairs suspend (also deletes paths) Failover to B volumes Recovery started at remote Local site returns Establish paths remote to local Failback B to A volumes Shutdown recovery FREEZE LSSs to suspend pairs Failover to local Establish paths local to remote Failback A to B volumes Start production Global Copy: Uses asynchronous processing to transmit data Global Copy is designed for those sites that conform to the following conditions: RPO of many hours or days Distances that may be > 300 KM Global Copy provides consistency by: Quiescing application at primary Allow the data to drain or go-to-sync FREEZE/RUN the pairs Consistency Group is NOT specified on establish path; therefore, extended long busy or I/O queue full not returned to server FlashCopy B volumes to C (C can be used for disaster recovery) Reestablish paths Resync the pairs Planned outage procedure same as forming consistency group Global Copy is a wonderful tool for data migration. Global Mirror Uses asynchronous processing to mirror data Global Mirror is designed for those sites that conform to the following conditions: RPO > 0 but still very current Distance that are greater than 300 KM Total of 8 disk systems in any combination of primary and secondary subsystems is supported. An RPQ can be requested to add more systems to the session. Global Mirror provides consistency by: Master session signals form consistency group Coordination time Creates Change Recording (CR) bit map and during this time, write I/Os are held Maintains order of dependent writes Once all CR bit maps are created, writes are once again responded to Drain Out of Sync (OOS) bit map to remote FlashCopy® secondary (B) volume to C volume Planned outage for testing: Pause Global Mirror session FREEZE/RUN to LSSs to suspend pairs Failover to B volumes Fast Reverse Restore (FRR) FlashCopy C>B Wait for background copy to complete FlashCopy B>C with the same options used during Global Mirror setup FlashCopy B>D Reestablish paths local to remote Failback A>B Resume session Test on D Global Mirror (continued) Unplanned outage: Terminate Global Mirror session if still active Issue Failover to B volumes Check FlashCopy status Revert or commit Fast Reverse Restore C>B Wait for background copy to complete FlashCopy B>C with same options used during initial Global Mirror setup Recovery can now be started on the B volumes Local site returns Establish paths from remote to local Failback with XD or gcp option to B volumes Once Global Copy is in steady state , quiesce host activity Insure that OOS=0 after quiescing host activity Failover to A volumes with XD or gcp option Establish paths from local to remote Failback to A volumes with XD or gcp option Reestablish any Master to Subordinate links that may have been lost Start the Global Mirror session Resume production activity to the A volumes CHAN EXT FC ‘C’ Target PPRC 'A' PRIMARY CHAN EXT PPRC ‘B’ Secondary and FC Source Global Copy over long distance FlashCopy Summary Global Mirror for zSeries (XRC) Metro Mirror Global Copy Global Mirror RPO > 0 (3-5 sec) > 0 (hours) > 0 (3-5 sec) Distance No Limit 300 KM (fcp) No limit Asynch Yes No CKD FB CKD & FB