© 2004 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice Business Copy.

© 2004 Hewlett-Packard Development Company, L.P.The information contained herein is subject to change without notice

Business Copy

Rev. 1 HP Confidential 2

Objectives

• Describe the similarities and differences between the snapshot copy and complete mirror copy methods of mirroring data

• Describe the BC pairs state transitions

• Diagram the BC pair tiers and label all Mirror Unit values for all volumes

• Describe the CV BC Swap and Freeze option

• Create multitier BC pairs in CV

• Explain the common BC terms

9 – 1


BC XP overview

• BC is an XP array feature that creates local mirror copies (not snapshots) of LUs within an XP disk array at hardware speeds, providing asynchronous internal data backup

• You can make up to nine copies of the data in an XP array

• Volume naming conventions are the same as for CA

– P-VOLs contain original data

– S-VOLs contain duplicate data

• BC performs asynchronous copies to S-VOLs, which results in a decrease in acknowledgement time and an increase in throughput from the host

• BC operations can be performed from

– SVP laptop in the XP disk array

– RM XP on the host with command line control

– CV XP

9 – 3


Factors that affect BC operation times

BC operation times depend on configuration

• Capacity of BC paired volumes

• P-VOL and S-VOL locations (same as different RAID groups, ACP pairs)

• HDD performance spindle speed

• RAID level (RAID1/RAID5)

• Time increases with increasing I/O load

• Copy Pace Parameter (slower/medium/faster)

9 – 4


P-VOLs and S-VOLs • When creating volume pairs, data on the P-VOL is copied to the S-VOL

to synchronize the volumes.

• P-VOLs normally remain available to all hosts for read/write I/O, except during pair resync-restore. Changes are copied from the S-VOL to the P-VOL during pair resync-restore.

• S-VOLs become available for host write access only after the pair is suspended.

• Read operations to the S-VOL are available except when reads are disabled with the paircreate command. (-m noread option) read does not make sense because data is “corrupted”

• While the BC volume pair is suspended, all changes to the P-VOL and S-VOL are logged in a track-level bitmap.

• Resync of the pair overlays on the S-VOL what changed during the suspend on the S-VOL plus what changed on the P-VOL (due to P-VOL and S-VOL updates).

9 – 4


Asynchronous write detail

The write operation is typical of cached arrays in which the I/O is acknowledged after it is in cache and then data is destaged asynchronously (and opportunistically) to disks.

9 – 5

XP disk array returns write complete to host as soon as data is written to cache memory

Data in cache memory is asynchronously written to both the P-VOL and S-VOL

Asynchronous write access to secondary volumes

Host

P -VOL

ACP pairACP pair

Maximum 9

S-VOLS

CacheCache memory

(1) Write I/O

(2) Write complete

Data

1

2

3

XP array

S-VOL

S-VOL

S-VOL


SMPL(simplex

)

COPY-PD(pending)

initial copy complete

PAIR

updatecopypaircreate

Business Copy XP Operations and States:Paircreate


PAIR COPY-SP(split

pending)pending copy complete

PSUSupdatecopy

pairsplit

Business Copy XP Operations and States: Pairsplit


Business Copy Fast Split

P-Vol

Asynchronous Copy

S-Vol

Read/Write Read-Only

Status: PSUS

P-Vol S-Vol

Read/Write Read/Write

Background Copy of Delta Data

Fast Split Complete

Data which isn’t copied to S-Vol yet.

Status: Pair

Split Request

Normal Operations S-Vol Access (e.g. for backup)

Split Times:

a few seconds!


Business Copy Fast SplitRAID Manager Changes XP1024/128

•no chance to choose quick or normal mode

•P-VOL immediately gets PSUS state, S-VOL gets COPY state

•pairevtwait cannot check state of S-VOL (SSUS) ---> old scripts implement new feature automatically use hand made scripting to check S-VOL state

•how to change array’s default behaviour:

mode 122 ON/checked pairsplit will work the old way mode 122 OFF/unchecked pairsplit will use fast Split


use environment variable $HORCC_SPLT

NORMAL

The "pairsplit" and "paircreate –split" will be performed as Non quick mode regardless of setting of the mode via SVP or Command View.

QUICK

The "pairsplit" and "paircreate –split" will be performed as Quick Split regardless of setting of the mode via SVP or Command View.

Business Copy Fast SplitRAID Manager Changes XP12000


Business Copy (BC) At-Time Split (1)

Applies to BC pairs that belong to a consistency group. • BC consistency (CT) group can ONLY be created during

paircreate process with RAID Manager XP interface

• Consistency (CT) groups are used with CA Async to time stamp I/Os to ensure concurrency and in order I/O delivery.

• CT groups (CTGs) ensure that when pairs are split, data in the Volumes is in a known state and can be used with data base logs.

• CV can perform operations on BC CT group pairs (At-Time Split configured pairs), but can not create the pairs (can split them, but not create them).

• 128 CTs on XP10247128 – 256 CTs on XP12000


At-Time Split functionality

Without At-Time Split

With At-Time Split

BC Split

P-Vol S-VolP-Vol S-Vol

P-Vol S-Vol

Written data time of split:

1, 2, 3, 4, 5

copied data time of split, without “At-Time Split”:

1, 3, 4,

BC At-Time Split

P-Vol S-VolP-Vol S-Vol

P-Vol S-Vol

Written data time of split:

1, 2, 3, 4, 5

copied data time of split, without “At-Time Split”:

1, 2, 3, 4, Consistency Group (CTG)


At-Time Split functionality

• keep in mind:− CT cannot be defined within CV – you can only see it there− you should use one CT for one logical group of LDEVs

within RAID manager− even with CT used, you can force to split a single pair

within the group− do not use different CTs within one logical group

• BC atomic split limits:− now 256 consistency groups instead of 128− up to 1024 BC pairs per consistency group− no mixture of Open System and Mainframe pairs

per consistency group


PSUS or

PSUECOPY-RS(resync)

resync complete PAIR updatecopy

pairresync

Business Copy XP Operations and States: Resync Pair


Business Copy Quick Restorenew scenario

what is quick RESTORE doing ?• swaps disk locations for P-VOL and S-VOL• starts normal „resync“ from P-VOL to S-VOL

4 resync options:• normal resync• fast resync• normal restore• fast restore


Business Copy Fast Resync/RestoreRAID Manager Changes XP1024/128

•no chance to choose quick or normal mode

•P-VOL immediately gets PAIR state, S-VOL gets COPY state

•pairevtwait will immediately get a PAIR state

•the P-VOL will be in the RCPY state at no time (not true for XP12000)

mode 80 ON/checked pairrestore will work the old way mode 80 OFF/unchecked pairrestore will use fast restore

(virtual restore)

mode 87 ON/checked pairresync will work the old way mode 87 OFF/unchecked pairresync will use fast mode


use environment variable $HORCC_RSYN

NORMAL

The "pairresync" will be performed as Non quick mode regardless of setting of the mode via SVP or Command View.

QUICK

The "pairresync" will be performed as Quick Resync regardless of setting of the mode via SVP or Command View.

Business Copy Fast Resync/RestoreRAID Manager Changes XP12000


use environment variable $HORCC_REST

NORMAL

The "pairresync -restore" will be performed as Non quick mode regardless of setting of the mode via SVP or Command View.

QUICK

The "pairresync -restore" will be performed as Quick Restore regardless of setting of the mode via SVP via Command View.

Business Copy Fast Resync/RestoreRAID Manager Changes XP12000


Some slide text here

P-Vol S-Vol

P-Vol S-Vol

P-Vol

P-Vol

S-Vol

S-Vol

•Oracle: start online backup alter tablespace X begin backup

•Oracle: stop online backup alter tablespace X end backup alter system switch logfile alter database backup controlfile vgchange -a y vgora mount /oradata ...

•XP: split BC pairsplit -g ORACLE

•XP: wait for SPLIT state pairevtwait -g ORACLE -s PSUS

•XP: reestablish mirror pairresync -g ORACLE

Backup concept with BC


Use CV to create BC pairs

Implement the XP Array BC pair creation process using CV XP• Show the paircreate task

• Create P-VOL/S-VOL pairs

• Discuss Mirror Unit numbers (MU#s)

9 – 10


BC high-level concept

9 – 11

Nine copy BC tiered configuration

P-VOL

S-VOL

S-VOL

S-VOL

S-VOL

S-VOL

S-VOL

S-VOL

Node volume

S-VOL

Node volume

S-VOL

Node volume

Root volume

Copy

Copy

Copy

Copy

Leafvolumes

Making duplicate volumes(BC pairs)

Making duplicate volumes(BC cascade volume pairs)

L1 pair

L2 pair

P-VOL: Primary volume S-VOL: Secoondary volume

Up to nine S-VOLs can becreated for each P-VOL

Java appletprogram

Web client Web server

CommandView PC SVP

UNIX/PC server

Disk Array


Layer concepts

• A volume can be designated an L1 volume, or an L2 volume, or both

• Tiers are another way to talk about the levels

– Tier 1 = Root volume

– Tier 2 = Node volumes

– Tier 3 = Leaf volumes

9 – 12-13

P-VOL

S-VOL/P-VOL

01

S-VOL/P-VOL

02

S-VOL/P-VOL

03

S-VOL04

S-VOL05

S-VOL06

S-VOL07

S-VOL08

S-VOL09

MU#=0

MU#=1

MU#=2

MU#=1

MU#=2

MU#=1

MU#=2

MU#=1

MU#=2

MU#=0

MU#=0

MU#=0

MU#=0

MU#=0

MU#=0

MU#=0

MU#=0

MU#=0

Root volume

Node volume

Leaf volumeL1 pair

L1 pair

L1 pair

L2 pair

L2 pair

L2 pair

L2 pair

L2 pair

L2 pair


MU#s

• The MU# is unique for a volume, but not for a pair — Both P-VOL and S-VOL could have an MU# of 0

• Every BC/CA volume has an MU#

• A blank MU# is implied to be 0, and usable for either BC or CA

• An explicit MU# of 0 means the volume is usable for BC only

• A volume can have an implied MU# of 0 and an explicit MU# 0

– As a BC P-VOL with an implied or explicit MU# = 0

– As a BC S-VOL with an implied or explicit MU# = 0

• Regarding BC

– MU# = 0 is the number used by a volume/LU to keep track of its first pair

– MU# = 1 is number used by a volume/LU to keep track of its second pair

– MU# = 2 is the number used by a volume/LU to keep track of its third pair

• the S-VOL side of a cascaded node volume, leaf volume or simple S-VOL uses allways MU# = 0

9 – 14


BC cascade notes:

• PAIRCREATE from root to leave ONLY

– with CV you have to do this in one step

– with RAID Manager you could do it one by one(and you could start with paircreate on L2 and add L1 pairs afterwards – will take longer for L1 than normal)

– you cannot create more than one cascade (CV & RM willnot allow this)

• PAIRSPLIT from root to leave ONLY

• no RESTORE is possible between L1 and L2

• in 1:n relationships on L1, restores are possible with all otherpairs on L1 being in PSUS state (steady and fast restore)


other notes:• in a PAIRSPLIT process (quick or normal) a PAIRSPLIT–S cannot

be issued before all delta has been moved to S-Vol monitor status of S-Vol moving from COPY to SSUS

• normally a PAIRSPLIT–S splits immediately while in PAIR state• what is PAIRSPLIT-E doing:

−marks complete S-Vol track table dirty

−will enforce a full copy for RESYNC

−will not allow a normal or quick RESTORE operation• in normal RESYNC, PAIRSPLIT cannot be issued before PAIR state• in normal RESTORE, PAIRSPLIT cannot be issued befor PAIR state• in quick RESYNC, PAIRSPLIT can be initiated at any time• in quick RESTORE, PAIRSPLIT can be initiated at any time, but you

have to monitor the status moving PSUSRCPPAIR


BC reservation• used as filter for CV when setting to reserved, change filter to

see LDEV• makes the potential S-VOL readonly• gets erased when the pair is finally deleted reservation is a one-time only flag


QUICK mode scenarios (1)Group PairVol(L/R) (Port#,TID, LU-M) ,Seq#,LDEV#.P/S,Status, % ,P-LDEV# M

geme geme_000(L) (CL1-B ,14, 5-0 )10278 21.SMPL ----,----- ----- -

geme geme_000(R) (CL1-B ,14, 6-0 )10278 22.SMPL ----,----- ----- -

# paircreate –g geme –split -vl

Group PairVol(L/R) (Port#,TID, LU-M) ,Seq#,LDEV#.P/S,Status, % ,P-LDEV# M

geme geme_000(L) (CL1-B ,14, 5-0 )10278 21.P-VOL PSUS, 100 22 W

geme geme_000(R) (CL1-B ,14, 6-0 )10278 22.S-VOL COPY, 0 21 -

# pairresync –g geme


geme geme_000(L) (CL1-B ,14, 5-0 )10278 21.P-VOL PAIR, 0 22 -

geme geme_000(R) (CL1-B ,14, 6-0 )10278 22.S-VOL PAIR, 0 21 -

# pairsplit -geme




...


geme geme_000(R) (CL1-B ,14, 6-0 )10278 22.S-VOL SSUS, 100 21 -


QUICK mode scenarios (2)# dd if=/dev/zero of=/dev/rdsk/c4t0d2 bs=64k count=100000

10000+0 records in

10000+0 records out

# pairdisplay -g geme -fcx



geme geme_000(R) (CL1-B ,14, 6-0 )10278 22.S-VOL SSUS, 73 21 W

# pairresync –g geme -restore


geme geme_000(L) (CL1-B ,14, 5-0 )10278 21.P-VOL RCPY, 73 22 -

geme geme_000(R) (CL1-B ,14, 6-0 )10278 22.S-VOL RCPY, 73 21 -

# pairevtwait –g geme –s pair –t 10000

pairevtwait : Wait status done.


geme geme_000(L) (CL1-B ,14, 5-0 )10278 21.P-VOL PAIR, 73 22 -

geme geme_000(R) (CL1-B ,14, 6-0 )10278 22.S-VOL PAIR, 73 21 –

# pairsplit –g geme




moves from PSUSRCPYPAIR


paircreate –m difP-VOL has content(1) at PAIRCREATE time

changing P-VOL after PAIRCREATE did reach PAIR state content(2) # pairdisplay -g geme -fcx


geme geme_000(L) (CL2-B ,18, 3-0 )10278 2f0.P-VOL PAIR, 99 2f1 -

geme geme_000(R) (CL2-B ,18, 4-0 )10278 2f1.S-VOL PAIR, 99 2f0 -


# pairdisplay -g geme -fcx


geme geme_000(L) (CL2-B ,18, 3-0 )10278 2f0.P-VOL PSUS, 100 2f1 W

geme geme_000(R) (CL2-B ,18, 4-0 )10278 2f1.S-VOL SSUS, 100 2f0 -

now P-VOL and S-VOL are identical –> content(2)

changing P-VOL (3) again will keep S-VOL unchanges (in SSUS state)

do a PAIRRESYNC results in (S-VOL still in content(2) state):# pairdisplay -g geme -fcx


geme geme_000(L) (CL2-B ,18, 3-0 )10278 2f0.P-VOL PAIR, 99 2f1 -

geme geme_000(R) (CL2-B ,18, 4-0 )10278 2f1.S-VOL PAIR, 99 2f0 –


now P-VOL and S-VOL are identical –> content(3)


paircreate –m dif•the S-VOL in PAIR state will not be updated – the S-VOL will either show:

−contents of P-VOL before PAIRCREATE of pair

−contents of P-VOL at PAIRSPLIT time, the split updates the S-VOLonce, the PAIRRESYNC will not do any further updates to the S-VOLif the P-VOL has been changed during split time


paircreate –m inc

P-VOL has content(1) at PAIRCREATE time

changing P-VOL after PAIRCREATE did reach PAIR state content(2)


CA / BC combinations

normal and quick RESTOREis possible



HORCTAKEOVERis possiblewith activeBC pairs



normal and quick RESTORE is possible



(4) normal OR quick RESTORE

(1) HORCTAKEOVER

(2) PAIRSPLIT

(3) PAIRSPLIT


BC states (1 of 2)

9 – 15

SMPL simplex Not currently paired

COPY (PD) pending Initial copy in progress

PAIR pair Initial copy complete, asynchronous updates are in progress

PSUS split Accepting write I/Os for P-VOL and S-VOL, maintaining different data on track maps

PSUS (SP) quick split Accepting write I/Os for P-VOL and S-VOL, differential map is being copies


BC states (2 of 2)

COPY (SP) suspend Updates issued prior to a split command are in progress

COPY (RS) resync P-VOL and S-VOL are being resynchronized, updates are in progress

COPY (RS-R)

reverse resync P-VOL and S-VOL are in process of reverse resync, updates are in progress

PSUE suspend Asynchronous updates to S-VOL have stopped, maps are not being maintained

9 – 15


BC transition table

9 – 15

P-VOL S-VOL

P-VOL S-VOL

SMPLPSUE

Not synchronized

Updated copy

P-VOL S-VOL

PAIR

(Initial copy completed)

Update copy

Initial copy

Split pair

Update copy (reverse)

Reservedvolume

Any pre-statusexcept PSUS(SP)

Pairsplit-S

Paircreate

Pairsplit Pairsplit

PSUS

Pairresync

Pairresync

P-VOL S-VOL

COPY(PD)

P-VOL S-VOL

Update copy

COPY(RS)

Pairresync

Pairsplit-E

P-VOL S-VOL

COPY(RS-R)

P-VOL S-VOL

COPY(SD) / PSUS(SP)

-Normal-Quick

-Normal-Quick

-Reverse -Quick restore


Swap and freeze functionality• Set in BC option area

• Alters the behavior of the Quick Restore Operation

9 – 16

With Swap and Freeze option

Without Swap and Freeze option

• attention:

– P-Vol and S-Vol will go into PAIRstate with a fixed minimum delta (may become bigger when P-Vol gets changed)

– cleanup requires a pairsplit/pairresync operation


Display filter menu• The Display Filter tab enables

various filtering options for the Display screen.

• If the Pair box is selected and non pair box is deselected, then only volumes in the Pair state display.

• This can be very useful to find and isolate volumes


BC option tab

You can set the BC options by selecting and/or clearing the option check boxes on the Option screen.


Paircreate Operation(1-4)


Paircreate Operation - Pair Create Window (2-4)

• The Paircreate window displays the P-VOL and S-VOL information for the pair(s) being created (added).• You can use the Paircreate window to select the S-VOL(s)for each P-VOL, set the copy pace for all pairs being created,and start the paircreate operation(s).


Paircreate Operation(3-4)

1

23

45

6


Paircreate Operation (4-4)

7

8


BC History screen

The History pane displays current BC pair status information as well as BC pair history information for the selected port. To open the History pane, click the History tab on the BC main screen.


Glossary (1 of 4)

Root volume• The volume that is the original P-VOL.

• The root P-VOL can participate in up to three L1 pairs.

Node volume• The S-VOL in relation to the root P-VOL.

• The node volume can also be an L2 P-VOL in relation to L2 S-VOL.

9 – 31


Glossary (2 of 4)L1 volume pair• The volume pair at the first layer

• It contains the root P-VOL and a node S-VOL

• Up to three L1 pairs can be created using the same P-VOL

• Each L1 S-VOL can be a P-VOL at the L2 layer for up to two L2 S-VOLs. This provides a total of 10 copies of the data (1 master and up to nine copies of the master)

L2 volume pair • The volume pair at the second layer

• This layer contains an L2 P-VOL, which is also an L1 S-VOL, as well as the L2 S-VOL

• The L2 S-VOL is also called the leaf

Leaf volume — This is the same as the L2 S-VOL

9 – 31


Glossary (3 of 4)

MU• Every volume can have up to three MU#s associated with it.

The MU# designates a pair for which the volume is a member. The MU# is not the number of the pair, but of the volume. A pair is made up of two volumes, of which each pair member could have the same or different MU#.– MU = 0 is the number of a volume for its first pair.

– MU = 1 is the number of the same volume for its next pair.

– MU = 2 is the number of the same volume for its third pair.

• The MU# of an S-VOL is always MU = 0.

9 – 32


Glossary (4 of 4)

Tier • Same as the level of a pair member

– Tier one would be the root volume

– Tier two would be the node volumes

– Tier three would be the leaf volumes

9 – 32


module end


Business Copy Performance

Initial Copy Speed within 1 ACPin MByte/sec

0,00100,00200,00300,00400,00500,00600,00700,00800,00900,00

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

#LDEVs, 72 GByte Drives

OPEN-V

OPEN-E



Initial Copy Speed between 2 ACPsin MByte/sec

0,00

100,00

200,00

300,00

400,00

500,00

600,00

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15


OPEN-V

OPEN-E



Resync Copy Speed within 1 ACPin MByte/sec

0,0050,00

100,00150,00200,00250,00300,00350,00400,00450,00

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16


OPEN-V

OPEN-E



Resync Copy Speed between 2 ACPsin MByte/sec

0,00

100,00

200,00

300,00

400,00

500,00

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15


OPEN-V

OPEN-E



Inital Copy Speed within 1 ACPin Mbyte/sec

0,00

100,00

200,00

300,00

400,00

500,00

600,00

700,00

1 2 3 4 5 6 7 8 9 10 11 12 13 14

-c Value, 8 LDEVs used

OPEN-V

OPEN-E

SVP copy pace parameter (slower – medium –faster) has no influence



Maximum BC Speed between 2 RGsin MByte/sec - 72 GByte Drives

0,00

50,00

100,00

150,00

200,00

250,00

300,00

350,00

1 2 3 4 5 6 7 8 9 10 11 12

initial OPEN-V

resync OPEN-V

© 2004 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice Business Copy.

Documents