user perspective of emerging emc technologies - gse...
TRANSCRIPT
User Perspective of Emerging EMC
Technologies
Session DH 2 November 2016
Jim Erdahl
U.S.Bank [email protected]
Lets Keep it Lite Mary and I have been married for more than 38 years.
Mary allows me to do everything I want to do,
as long as I ALWAYS do what she tells me to do.
Mary’s love for me causes her to shop a lot.
We SCUBA so that we don’t need to hear each
other speak.
U.S. Bank and EMC
U.S. Bank and EMC have worked together to
introduce:
• Host Read Only (HRO) Devices USA SHARE-March 2016
• Universal Data Consistency™ USA SHARE-March 2014
• Data Protector for z Systems (zDP) Today
Like a Kid in a Candy Store…
Who got to go to the factory and eat off the line!
Agenda zBoost™
– Performance
– zHPF
– PAV Optimizer
VMAX V3 Beta – Hardware
– MFE 8.0
GDDR 5.0 Beta
DC3: V3 Cutover
zBoost™
• No charge* non-disruptive microcode upgrade to
improve performance for Mainframe VMAX20K
and VMAX40K.
– Improves the throughput with FICON
– Provides full zHPF support
– Enhances the use of PAV’s
*Support for zHPF is a chargeable feature on VMAX, part of the Mainframe Essentials bundle.
zBoost Performance Benefits (VMAX 40K)
• Improves the maximum IOPS by moving a back end processor core to the front end for FICON (previously known as Mainframe Performance Accelerator - MPA)
– Increases usable front end director capacity
– Improved IOPS for FICON
– Reduced response time
– Batch run-times were improved
zHPF Support
• Verify hardware support – D IOS,ZHPF
• We have not performed
any measurements
• Worth Mentioning – No issues with our z13 Processors
– On z12, non-disruptive Channel Detected Errors
– Heard that the FICON fiber needs to be clean
zBoost™ PAV Optimizer Objectives
• Significantly improves performance of
multi-track zHPF I/O
• Extend parallel processing programming
paradigms into the I/O itself
• Ensure transparent implementation for
ease of exploitation
– No JCL or program changes necessary
zHPF I/O optimized with PAV Optimizer
Tracks 4-6 Complete
z/OS IOS
Subsystem
zHPF Channel Pgm
Read 9 tracks
Device 1901
TCW
TCCB
DCW
READ TRACK
Count=9
A
c
C
e
S
S
M
e
t
h
o
d 19F5
Alias
19F1
Alias
1901
Base
P
A
V
O
P
T
I
M
I
Z
E
R
PAVO • 64 HyperPAV’s per 192 devices-already heavily
used
• Early use of the functions – Beta via Early Ship
– 16 Jobs assisted since late Oct 2015
– > 1,100 devices
Project was delayed because ………
Sample SCF INI PAVO Parms:
SCF.LFC.LCODES.LIST=1234-5678-9012-3456 /* PAV OPT */ ******************************************************** /* PAV OPTIMIZER - ENABLEMENT SCF.DEV.OPTIMIZE.ENABLE=YES SCF.DEV.OPTIMIZE.PAV=YES /* YES/PASSIVE */ SCF.DEV.OPTIMIZE.SMF.RECID=203 /* SMF RECORD ID NUMBER */ /* SCF.MSC.PAVO=YES /* SUSPEND/RESUME AT CYCLE SWITCH */ /* /* PAV OPTIMIZER - SELECTION (SITE SPECIFIC) SCF.DEV.OPTIMIZE.PAV.STORGRP.INCLUDE.LIST=SGONE,SGTWO SCF.DEV.OPTIMIZE.PAV.JOBPREFIX.LIST=JAERDAH SCF.DEV.OPTIMIZE.PAV.JOBNAME.LIST=JOBNAME1,JOBNAME2 /* /* PAV OPTIMIZER - RESOURCE UTILIZATION SCF.DEV.OPTIMIZE.PAV.TRACK.MIN=2 SCF.DEV.OPTIMIZE.PAV.SPLIT.MAX=8 SCF.DEV.OPTIMIZE.PAV.SPLIT.MAX.READ=4 SCF.DEV.OPTIMIZE.PAV.SPLIT.MAX.WRITE=8 SCF.DEV.OPTIMIZE.PAV.QUIPOINT.GLOBAL=5000 SCF.DEV.OPTIMIZE.PAV.QUIPOINT.LCU.PCT=75 SCF.DEV.OPTIMIZE.PAV.QUIPOINT.DEVICE=8
As provided, do not
use
Selection Criteria for Batch Job Optimization • Mission critical Application
• A Lot of zHPF I/O (SMF Type 42)
• Significant Run Time Reduction
• Job is in the critical path
• Under MBOS* influence (larger buffers)
*Mainview Batch Optimizer from BMC
PAVO Results
Note: These are development jobs, so the data tested with varies greatly.
-10.0
0.0
10.0
20.0
30.0
40.0
50.0
60.0
Percent Runtime Reduction
Percent Runtime Reduction
What’s Next:
We have engaged EMC resources
We need to better understand how to use the SMF
data
We have requested enhancements
• Selection by data set name
• Need passive and active mode in parallel
VMAX V3 Beta Validation
• Hardware – Mainframe Engines
• Mainframe specific microcode
• Mainframe Enabler – MFE Version 8.0
• Geographically Dispersed Disaster Restart – GDDR Version 5.0
The objective was to validate functions, not performance
Initial Beta Configuration
.
3 LPARS
RA00
RB00
R900
VMAX3
DC1 DC2
z13
SRDF/A • VMAX 200K
• 4TB
• 8 FICON
VMAX3
VMAX V3 Beta
• The Beta validation was a success. This is what we have participated with:
– I/O Driver provided by EMC and with our software
– SRDF/A with MSC and MCM
– TF Mirror Clone Emulation
– SnapVX
– Batch zDP
– Host Read Only
– ISPF Interface for zDP
– SRDF/S
– ConGrp
– Autoswap
– Unisphere
Beta Validation Process • Education – Train the trainers
• At each phase, delivery included:
– Microcode
– MFE 8.0 Software
– New/updates to manuals (w/o messages)
– Validation Script
– Validation JCL
• Enhanced Validation
SnapV Method • Historical TimeFinder technologies
– Need a target device at the time of the Copy
Copy to target
Source Volume Target Volume
Source Volume Target Volume
Copy to target
This slide was originally prepared by Justin Bastin, EMC
TimeFinder SnapVX
• Built on Thin provisioning – Data stored in thin pools (SRP)
• Volume level only
• Supports up to 256 ‘target-less’
snapshots per volume
• Single architecture supporting: – TF/Mirror (via Clone Emulation)
– TF/Clone
– TF/SNAP
– SnapVX
This slide was originally prepared by Justin Bastin, EMC
Production
Volume
Linked Target
Snapshot
Snapshot
Snapshot
Storage
Resource
Pool (SRP)
TimeFinder SnapVX • New SnapVX commands
– CREATE – Create snapshot structure of source with unique name
– ACTIVATE – Obtain point it time
– LINK/UNLINK – Associate/disassociate PiT snapshot to target volume
– QUERY – Displays information on the snapshot
– RENAME – Change name of snapshot
– TERMINATE – End snapshot once no longer linked
TimeFinder SnapVX– Pointer-Based SNAPs
12 Noon
Source Device
(TDEV)
Snapshot (up to 256)
10 A.M.
Target Device(s)
(TDEV)
Storage
Resource
Pool
CREATE & ACTIVATE LINK / UNLINK
TERMINATE
Host volume
Snapshot
Pointer
Data
structure
This slide was originally prepared by Justin Bastin, EMC
zDP Uses SnapVX Data Protector for z Systems (zDP) delivers the capability to recover from logical data corruption with minimal data loss.
Yes, it was conceived in a bar
But, professionally designed ….
zDP - How does it work? • Built on top of SnapVX
• Maintains Consistency of data across volumes and VMAX V3’s, using either ECA or SRDF
• Creates up to 256 Snap Sets every 10 minutes – (42.5 hours)
• Retain Snap Sets
• Terminate Snap Sets
• EMC has a tool to size your SRP using ChangeTracker
zDP Validation • First - Performed the validation as requested from EMC with
two volumes
• Second - Enhanced validation was with: – 94 volumes (MOD 3, 9, 27, 54, & EAV’s)
– 10 minute cycle time
– Ran I/O Driver process generating about 200 writes per second
• Results – Ran for more than week
– Used less than 34GB of the SRP
Operator Commands • Start: F emcscf,ZDP,START vdg_name
• Stop: F emcscf,ZDP,STOP vdg_name
• Pause: F emcscf,ZDP,PAUSE vdg_name
• Resume: F emcscf,ZDP,RESUME vdg_name
• Locks: F emcscf,ZDP,RELEASEDEVICELOCK vdg_name
vdg_name is case sensitive
The zDP Process
• Define the Versioned Data Group (VDG)
• Add the source volumes to the group
• Define the Target Set (TGT)
• Add the target devices
• Start the zDP process
• Celebrate – that was easy
zDP Batch JCL
//STEPA EXEC PGM=EIPINIT,REGION=0M
//STEPLIB DD DISP=SHR,DSN=SYS2.EMC.MFE800.LINKLIB
//SYSPRINT DD SYSOUT=*
//SYSUDUMP DD SYSOUT=*
//SCF$EMCL DD DUMMY ---> YOUR EMCSCF ADDRESS SPACE
Statement1
Statement2
….
Statement(n)
/*
zDP Define Versioned Data Group GLOBAL MAX_RC(4)
DEFINE VDG VDGR900A,
CYCLE_TIME(10,0),CYCLE_OVERFLOW(NEXT),
CONSISTENT(YES),TIMEOUT(30),
TERM_POLICY(OLDEST),
SRP_WARN%(80),
MAX_SNAPSETS(255), Save 1 snapset for other SnapVx
SAVED_SNAPSETS(1,3),
PRESERVED_COPY_LIMIT(003),
LOG_OPT(SCF),SMFREC(204,VOLUMES),
EXIT(NONE),MAXRC(4)
MODIFY VDG VDGR900A,ADD,SYMDEV(FD2C,00A4-00AF)
MODIFY VDG VDGR900A,ADD,SYMDEV(FD2C,00B0-00BE)
MODIFY VDG VDGR900A,ADD,SYMDEV(FD2C,046C-048B))
zDP Define Target Group GLOBAL MAX_RC(4)
DEFINE TARGET_SET TGTR900A
MODIFY TGT TGTR900A,ADD,SYMDEV(FD2C,06E0-06EB) MODIFY TGT TGTR900A,ADD,SYMDEV(FD2C,06EC-06FA) MODIFY TGT TGTR900A,ADD,SYMDEV(FD2C,0AA8-0AC7)
Let’s fire this baby up !
zDP Start Command F emcscf,ZDP START vdg_name
14.41.36 STC23778 SCF0740I ZDP START VDGR900A
14.41.36 STC23778 SCF0741I ZDP START command accepted
14.41.36 STC23778 SCF0746I ZDP VDG VDGR900A Started
14.41.36 STC23778 EIP0200I *** EMC zDP - V8.0.0 (000) - Friday, February 26, 2016 ***
14.41.39 STC23778 EIP0201I VDG VDGR900A, Beginning cycle 1, Snapset VDGR900A.......160571441S00001
14.41.40 STC23778 EIP0217I VDG VDGR900A, Devices validated for consistency, via SRDF/A
14.41.43 STC23778 EIP0204I VDG VDGR900A, Snapset VDGR900A.......160571441S00001 created
14.41.45 STC23778 EIP0202I VDG VDGR900A, Completed cycle 1, next cycle scheduled for 14:51:39
14.41.53 STC23778 SCF1301I MSC - TASK TIMER
14.46.53 STC23778 SCF1301I MSC - TASK TIMER
14.51.39 STC23778 EIP0201I VDG VDGR900A, Beginning cycle 2, Snapset VDGR900A.......160571451C00002
14.51.40 STC23778 EIP0217I VDG VDGR900A, Devices validated for consistency, via SRDF/A
14.51.44 STC23778 EIP0204I VDG VDGR900A, Snapset VDGR900A.......160571451C00002 created
14.51.46 STC23778 EIP0202I VDG VDGR900A, Completed cycle 2, next cycle scheduled for 15:01:39
Link a Snap Set LINK VDG(VDGR900A),SNAPSET(160571501C00003) TGT(TGTR900A)
EIP0001I *** EMC zDP - V8.0.0 (000) - SCF V8.0.0 (000) *** 07:30:47 02/28/2016
EMCP001I LINK VDG(VDGR900A),SNAPSET(160571501C00003) TGT(TGTR900A)
EIP0053I SYMM FD2C/0001967-01562, Linking SNAPSET VDGR900A.......160571501C00003
EIP0034I LINK command completed
EIP0002I All control statements processed, highest RC 00
Keep a Snap Set PERSISTENT SET,VDG(VDGR900A),SNAPSET(160571451C00002)
EIP0001I *** EMC zDP - V8.0.0 (000) - SCF V8.0.0 (000) *** 06:44:48 02/27/2016
EMCP001I PERSISTENT SET,VDG(VDGR900A),SNAPSET(160571451C00002)
EIP0060I SYMM FD2C/0001967-01562, PERSISTENT SET for SNAPSET VDGR900A.......160571451C00002
EIP0034I PERSISTENT command completed
EIP0002I All control statements processed, highest RC 00
zDP QUERY Commands
Versioned Data Group Query QUERY VDG VDGR900A,STATUS
QUERY VDG VDGR900A,DEVICE
QUERY VDG VDGR900A,SNAPSET
QUERY VDG VDGR900A,SNAPSET,DETAIL
Target Query QUERY TGT TGTR900A,STATUS
QUERY TGT TGTR900A,DEVICE
QUERY TGT TGTR900A,STATUS,DEVICE
zDP Query Command of Snapsets
EIP0001I *** EMC zDP - V8.0.0 (000) - SCF V8.0.0 (000) *** 07:31:09 02/28/2016 Page 1
EMCP001I QUERY VDG VDGR900A,SNAPSET
EIP0035I Snapset Query for VDG VDGR900A
EIP0023I SYMM 0001967-01562, Microcode level 5977_0799, Type VMAX200K
EIP0024I Gatekeeper FD2C, Device Count: 59
EIP0025I SRP ID/Name: 0001/SRP_1, Reserved Capacity: 10%
EIP0026I Total Capacity: 3359M, Total Allocated: 16M, Snap Allocated: 487
EIP0036I CREATE SOURCE_TRACKS EXPIRATION
EIP0036I SNAPSET_NAME STATE DATE TIME CHANGED UNIQUE DATE TIME
EIP0036I ____________________________ _____ ___________________ ______________ ___________________
EIP0039I VDGR900A.......160571441S00001 ACT-S 02/26/2016 14:41:41 13485 79 02/29/2016 14:41:40
EIP0039I VDGR900A.......160571451C00002 ACT-P 02/26/2016 14:51:41 13477 53
EIP0039I VDGR900A.......160571501C00003 ACT 02/26/2016 15:01:40 13164 53
………..
EIP0039I VDGR900A.......160590721C00245 ACT 02/28/2016 07:21:42 83 52
EIP0039I VDGR900A.......160571501C00003 LNK 02/28/2016 07:30:54 13164 53
EIP0034I QUERY command completed
zDP Device Locks F emcscf,ZDP,RELEASEDEVICELOCK, vdg_name
07.31.10 STC03576 SCF0740I ZDP,RELEASEDEVICELOCK VDG_A_1
07.31.10 STC03576 SCF0741I ZDP RELDLOCK command accepted
07.31.10 STC03576 SCF0746I ZDP VDG VDG_A_1 Started
07.31.10 STC03576 EIP0200I *** EMC zDP - V8.0.1 (027) - Thursday, July 28, 2016 ***
07.31.39 STC03576 EIP0203I VDG VDG_A_1, Ended - Thursday, July 28, 2016 ***
07.31.39 STC03576 SCF0747I ZDP VDG VDG_A_1 Ended
Lessons Learned • All volumes in a Snap Set must be in the same state
– SRDF/A
– ECA
• Target volume can not be smaller than the source – The define of Target Group works, but…….
– LINK command will verify sizes, then fail
• I prefer batch setup over ISPF (I’m a CSECT guy)
• At this time, GDDR does not play with zDP
GDDR V5.0 Validation – 2 Site • Still using two VMAX V3
• Much like the MFE 8.0 Validation
• Software libraries delivered in XMIT format
• Converted our GDDR Config 3 site to 2-site with Autoswap
• Several Planned Autoswaps
• Performed an Un-Planned Autoswap
GDDR 5.0 Beta Configuration
DC1 DC2
DC3
4 LPARS
RA00- DC1 C-System
RB00- DC2 C-System
RC00- DC3 C-System
R900- Managed System
IBM z13
SRDF/S
R900 Uses Subchannel Sets
SS2 to DC1
SS0 to DC2
SS3 to DC3 VMAX V3 & DLm
VMAX V3 & DLm VMAX SE & DLm
S
S
3
GDDR 5.0 Notes • To convert, copy GDDR V4.1 backup into V5.0
• New RACF Rules
• Additions to IKJTSOxx
• E05TFDEV and E04SRDFA members
• Refresh LMOD GDDRXG1A into dynamic LPA
• Not able to mix TF/Mirror (PGM=TMCTF) with TF/Clone or TF/SnapVX
(PGM=EMCSNAP) for consistency function
– GDDR 4.2 used TF/Mirror exclusively
– GDDR 5.0 adds support for TF/Clone and TF/SnapVX • Users with mixed VMAX2 and VMAX3 must migrate to TF/Clone to allow common consistency function
• GDDR supports common consistency across V2 (TF/Clone – Snap Volume) and V3 (TF/SnapVX –
CREATE) with a single ACTIVATE command in EMCSNAP
DC3 VMAX V3 Cutover Process Prepare:
Recruited Application Test Teams to Validate
Generated 72 EMC I/O Driver Jobs
Stage V3 Hardware & Pray that 100 cables are correct
Updated MFE to 8.0 and GDDR to 5.0 o GA Base code released on March 30, 2016
Converted DLm VMAX V2 from TF/Mirror to SNAPV o Copy all the data
o Dynamic QOS reduced run time from est. of 19 days to 7 ½
DC3 VMAX v3 Cutover Process Execution:
Initialized SA-Dump volumes with ICKDSF
Defined SRDF and pushed the data from DC1
Swapped the cables to the V3’s
Converted GDDR to 2-site w/Autoswap
Stopped SRDF/A and deleted all definitions
DC3 VMAX v3 Cutover Process Execution (Continued): Monitored SRDF progress with PGMITRKA
(from Lam Hairston)
Activated MSC
Converted GDDR back to 3-site w/Autoswap
DC3 VMAX v3 validation
Engaged Application Teams (friends) to validate
Ran our 72 EMC I/O Driver jobs
Brought up 13 systems, z/OS 2.1 & z/OS 2.2
Activated 10 day CBU
Cutover issues Who would believe me if I said there were none ?
Our Network Bandwidth between DC1 and DC3
Three Bin’s needed to be changed
o The wrong file was installed
o We did not communicate correctly to EMC
No Sev-1 Issues Created
Better Make a Dentist Appointment
Thank You for letting me share my trip through the
Candy Store and then through the Candy Factory
