66. GSE zos Expertenforum 2007 GDPS z/os Global Mirror Versus Global Mirror walter.muehlenstaedt@ch.ibm.com and System z9 2006 IBM Corporation
Agenda 1. Business Continuity Overview Business Continuity Objectives 2. Unlimited Distance D/R Solution Global Mirror z/os Global Mirror 3. CA/DR Solution (3 sites) z/os data only z/os and Open data 2
Business Continuance Objectives Determine your Objectives for Business Continuance (by application) Recovery Time Objective (RTO) how long can you afford to be without your systems? Recovery Point Objective (RPO) Data loss YES /NO when it is recovered, how much data can you afford to recreate? Network Recovery Objective (NRO) how long to switch over network? Determine cost / recovery time curve If I spend a little more, how much faster is Disaster Recovery? If I spend a little less, how much slower is Disaster Recovery? RPO Backup Application Processing Application Failure Determining the cost vs. RTO recovery curve is the key to selecting proper solution(s) 3
z/os Global Mirror Productivity tool that integrates management of XRC and FlashCopy with GDPS Full-screen interface Invoke scripted procedures from panels or through exit GDPS/XRC runs in the SDM location and interacts with SDM(s) Manages availability of SDM Sysplex Execute Scripts to perform a fully automated site failover Single point of control for multiple / coupled Data Movers production systems primary disk subsystems secondary disk subsystems SDM systems GDPS /XRC Flash Copy Devices State, Journal, Master, Control- Files 4
z/os Global Mirror XRC Update Processing Input Queue - IO's queued to read from primary CU XCGCO - Consistency groups queued to XLOPE Q XJTSK - Journal Queue XLOPE - Consistency groups queue to secondary's Residual count - Percentage of side file Limit based on HDS mode settings J1 A B J8 C A' B' 5
z/os Global Mirror XRC Update Processing 1. Timestamped application write 3. SDM requests data transfer 5. SDM forms consistency group 4. Updates sent to SDM 6. SDM journals consistency group 7. SDM writes consistency group to secondary vols 2. Primary disk subsystem maintains updates in cache J1 8. SDM notes consistency group complete A B J8 C A' B' 6
z/os Global Mirror Hlq.XCOPY.PARMLIB Also SYS1.PARMLIB(ANTXIN00) Contains tuning parameters Member name ALL or sessid Read at XSTART time Verified and changed with XSET command Hlq.XCOPY.msessid.MASTER Dataset for coupled SDM s 1 Cylinder Supports 14 SDM s Used to coordinate consistency between SDM s Hlq.XCOPY.sessid.CONTROL Sequential dataset Keep track of journal & secondary updates Read at XRECOVER to apply latest updates, ensure consistency Typically placed on same volume as MASTER dataset in coupled environment 7
z/os Global Mirror Hlq.xcopy.sessid.jrnl01 16 Used in odd even pairs Striped 2 stripes Should be placed in odd even LSS s Best performance if not on secondary CU (not always practical) Hlq.XCOPY.sessid.STATE PDSE Read at XRECOVER to relabel secondary volumes Typically placed on same volume as MASTER and CONTROL dataset MONITOR1 Records here Used by XRC Monitor 8
66. GSE zos Expertenforum 2007 Global Mirror Overview 2006 IBM Corporation
DS 8000 DS 6000 ESS 800 ESS 750 Native Performance Master Control Server LOCAL Global Mirror One of many possible configurations SAN Fabric Performance Transmission REMOTE FlashCopy DS 8000 ESS 750 ESS 800 Turbo Consistent Data Designed to Provide: Global Distance: Two-site, unlimited distance, data consistent asynchronous disk mirroring Heterogeneous: Data can span zseries and open systems data, and can contain a mix of zseries and open systems data Scalability: Consistency Group supported across up to 17 total ESSs in Global Mirror session (with RPQ) Flexibility: Many possible configurations Application Performance: Native Mirroring Performance: Two ESS Fiber Channel disk mirroring links per ESS sufficient for almost all workloads Intended Benefits Autonomic: No active external controlling software required to form consistency groups Saves cost: No server cycles required to manage consistency groups Lowers TCO: designed to provide improved performance, global distances, and lower costs 10
DSS LSS Global Mirror MASTER Session number A 12 A SBINFO START PAUSE RESUME STOP DEFINE/ UNDEFINE JOIN/ REMOVE SUBORDINATE A 12 11
Global Mirror A B C A Establish PPRC Paths Establish PPRC (XD) Pairs (A to B) Establish FC (ASYNC) (B to C) Done via the PPRC link (in-band) Establish SESSIONLINK to Subordinate Define SESSION to each primary LSS JOIN volumes to the session (within the LSS) START GM session with the MASTER CG creation starts - if all initial copy is done No action in Recovery Site! 12
Global Mirror Coordination time COORDINTERVAL CGINTERVAL DRAIN to B disk FlashCopy Coordination CGDRAIN COORDINTERVAL (ms def=15ms) and CGDRAIN (sec) is the max allowed time If exceeded, the CG will be lost but retried next time around! RPO will be CGINTERVAL (sec) plus the time to form a copy, so it may vary over time 13
10 9 8 zp2 GDPS/Global Mirror Site 1 Failure 12 11 1 12 11 1 Blue sysplex Red sysplex Non sysplex 7 6 5 2 4 3 Application Site CF1 zp3 zp5 zp1 zp4 Open Systems S S S S S L O7 CF2 10 9 8 7 6 GDPS K-sys O6 z/os and Open Systems sharing disk subsystem 5 2 4 3 GDPS R-sys NetView communication Recovery Site z/os and Open Systems sharing disk subsystem Capacity Back Up (CBU) Discretionary Backup open systems Global Mirror over Unlimited Distance S S S S S L Application site can have single z/os Systems, Open Systems, Systems in a Sysplex All data (z/os and Open Systems) can be mirrored using Global Mirror K-sys activities Manages multiple Global Mirror sessions Sends device info, scripts, alerts to R-sys R-sys activities: Secondary disk recovery, CBU activation, activate backup LPARs, IPLs systems. 14
66. GSE zos Expertenforum 2007 CA/DR Solution (3 sites) Differences 2006 IBM Corporation
A z/os Global Mirror PPRC C B K R SDM XRC PPRC Hyperswap In GDPS V3.3 a planned Hyperswap will not work Execute a unplanned Hyperswap Switch from A to B XRC Sessions from a to C are unusable Reestablish the XRC Session s (Initial Copy!) If downtime from A takes longer you must decide to establish XRC from B If downtime from A short then you can wait but this is a Business decision 16
z/os Global Mirror zseries Solution metropolitan unlimited Production Site 1 Site 2 Site 3 distance distance CF CF Parallel Sysplex CF Parallel Sysplex CF P' PPRC secondary GDPS/ PPRC P X PPRC primary XRC primary XRC X' F' XRC secondary Continuous Availability GDPS PPRC or GDPS/PPRC HM Designed to provide continuous availability and no data loss between sites 1 and 2 Sites 1 and 2 can be same building or campus distance to minimize performance impact Site 2 servers optional Site 1 failure Disaster/Recovery ƒ Switch to Site 2 disk (if server exists on Site 2) Site 2 failure ƒ Production continues in Site 1 Site 1 and 2 failure ƒ Failover to Site 3 with minimal data loss Continuous Availability, No data loss, Unlimited Distance 17
Metro & Global Mirror zseries and Open Solution metropolitan unlimited Production Site 1 Site 2 Site 3 distance distance CF CF Parallel Sysplex CF Parallel Sysplex CF P' PPRC primaries GDPS/ PPRC P X PPRC secondaries GM primaries Global Mirror X' F GM secondaries Continuous Availability GDPS/PPRC Designed to provide continuous availability and no data loss between sites 1 and 2 Sites 1 and 2 can be same building or campus distance to minimize performance impact Site 2 servers optional Site 1 failure Disaster/Recovery ƒ Switch to Site 2 disk (if server exists on Site 2) Site 2 failure ƒ Production continues in Site 1 Site 1 and 2 failure ƒ Failover to Site 3 with minimal data loss GDPS Managed coordinated solution for zseries and open systems 18
Questions 19