IBM Presentations: Smart Planet Template

advertisement
Chuck Laing- Senior Technical Staff Member (STSM)
22 October 2013
Data and Storage Migration Methods for
Power Systems
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
http://ibmtechu.com/ent
pST541
2
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Objectives
 Understand the underlying disk structure of LUNs
 Understand the virtual nature of SVC, DS8K, V7000 and XIV Storage
 Considerations that trigger migration
 Key decision factors
 Concepts for migration (a 4 step migration process)
 My object is to teach:
– Concepts that will make you better
– Concepts that will make your job easier
© 2013 IBM Corporation
3
3
Data and Storage Migration Methods for Power Systems pST541
Addressing data migration issues
 How do I address and avoid:
– Extended or unexpected downtime
– Data corruption
– Missing data or data loss
– Application performance issues
– Technical compatibility issues
4
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Knowledge is POWER!
 As System’s Administrators – we don’t always KNOW what we
don’t know about storage
– Ask for storage, leveraging what you know
– Avoid bottlenecks
– Use tools available
– Speed problem isolation
– Make more informed architectural decisions
 As Storage Administrators – we don’t always KNOW how the
storage will be utilized
– Make more informed architectural decisions
– Ask what is needed for best performance and IO separation
 What we are NOT going to do today:
– Try to turn Sys Admins into Storage Admins or vice versa
– Boil the ocean
5
© 2013 IBM Corporation
5
Data and Storage Migration Methods for Power Systems pST541
Problem statement
 Storage and System Administrators often clash in the common goal to
achieve data performance and availability, leading to:
– Too many logical configuration related outages
– Performance related enhancements not working to specification
– Leading causes:
• Lack of understanding configurations
• No cohesiveness between the logical and physical implementations
• Lack of communication between System and Storage Administrators
 Resulting in:
– A lack of data reliability and IO throughput
6
6
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
The top ten things to know
Systems Administrators should
know about Storage
1. What should I be aware of/what should
I avoid? (Tips & Pitfalls-Tuning)
2. Storage Overview - what's inside?
– What is the Physical makeup?
– What is the Virtual makeup (good
throughput design tips)?
– What is a Storage Pool ?
– Where do I place data?
3. Connectivity- Picking the right drivers
4. Host Attachment Kits
5. How to Improve Performance using
LVM
6. Documentation - why it matters
7. Topology Diagrams
8. Disk Mapping (view at a glance)
9. Easy Storage Inquiry Tools
10. Bottlenecks
7
© 2013 IBM Corporation
Storage Admins should know
about Logical Volume Manager
(LVM)
1. What should I be aware of/what should
I avoid? (Tips & Pitfalls-Tuning)
2. Hdisk Volume (LUN) purpose ?
– DB type
– Access Patterns
– Number of spindles required
– Stripe
– Spread
– Mirror
1. What is a Volume Group (VG)?
2. What I a Logical Volume (LV)?
3. Disk Mapping (view at a glance)
4. Easy Storage Inquiry Tools
5. Bottlenecks
6. ………………………
7. ………………………..
Data and Storage Migration Methods for Power Systems pST541
Summary


Knowing - what's inside will help you make informed decisions?
You should make a list of the things you don’t know
– Talk to the Storage Administrator or those who do know

A better Admin understands
–
The backend physical makeup
–
The backend virtual makeup
–
What's in a Storage Pool for better data placement
–
Avoids the Pitfalls associated with IO Tuning
Knows where to go to get right device drivers
Knows why documentation matters
–
Keeps Topology Diagrams
–
Keeps Disk Mapping documentation
–
Is able to use Storage Inquiry Tools to find answers
Understands how to troubleshoot storage performance bottlenecks



8
© 2013 IBM Corporation
8
Data and Storage Migration Methods for Power Systems pST541
A four step migration process
Evaluate
Validate
Migration Process
Execute
9
© 2013 IBM Corporation
Plan
Data and Storage Migration Methods for Power Systems
Data and Storage Migration Methods for
Power Systems
Evaluate
10
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Evaluate the data migration process
 Migrating data is always a disruptive process. Whatever the migration
technique used, it always affects to some degree the normal operations of
the system.
– Selecting the appropriate technique depends on:
• The criticality of the data being moved
• The resources available
• Other business constraints and requirements.
 Note: Risks should be identified depending on the migration technique used. We strongly
recommend that you consider selecting the technique that is the best compromise between
efficiency and the least impact to the system users.
11
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Evaluate the migration technique summary
 Make a list of Pros and cons (each offering strengths and limitations)
Migration technique
Pros
Cons
– Host-based
– Generally lowest initial implementation
cost
– Leverages existing and IP network
– LVM or LDM tools available
– Storage device-agnostic
– Leverages existing Operating System
skills
– Migration can happen on-line during peak
hours
– Consumes host resources
– Operating system specific
– Management can become complex and
time consuming
– Each host is its own island – no central
management console
– May cause an initial outage to install the
utility or software if it is not already
existing on the host
– Supports heterogeneous environments –
servers and storage
– Single point of management for replication
services
– Higher initial cost due to hardware &
replication software
– Requires proprietary hardware and may
require implementation of Storage
– Migration can happen on-line during peak
hours
– Supports heterogeneous environments –
servers and storage
– Single point of management for migration
– Requires an initial outage to bring the
host volumes on-line to SVC
– Requires the host to reboot to load or
upgrade the multipathing drivers
– Does not require additional special tools,
software or hardware
– Does not require additional skills or
training
– Requires the disruption of the
applications and down time
– Slow and cumbersome
– LVM
– LDM
– Add-on software such
as VxVM
– Volume (block) level
– TDMF
– Network-based
– Fabric
– TDMF-IP
– Application-based
– SVC
– Tape based
– TSM
– Etc
12
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Evaluating key decision factors
Key Factors
Description
Capability
– Performance
How quickly can data be copied from the
source to the target, balanced against system
overhead
–
–
TDMF
SVC
– Primary volume/
Source data
protection
If something goes wrong, the migration can be
terminated and application processing
restarted or continued on the source
data/device
–
–
–
–
TDMF
SVC with limitations
LVM / LDM
Tape based
– Implement tiered
storage
Moving data to a different array or to different
storage media for cost performance without
disruption
–
–
–
TDMF
SVC
LVM / LDM
– Multi-vendor
environments
Many data centers use hardware from several
vendors, which can result in source and target
hardware utilizing different vendors tools
–
–
–
–
–
TDMF
SVC
LVM / LDM with
possible restrictions
Fabric
Tape based
–
–
–
–
–
All with limits
TDMF
SVC
LVM / LDM
Fabric
– Application downtime
13
© 2013 IBM Corporation
Applications have different levels of business
criticality and therefore have varying degrees
of acceptable downtime
Data and Storage Migration Methods for Power Systems pST541
Evaluating migration triggers
14
Disk consolidation can trigger data
migration of storage when:
Storage migration can trigger LVM
data movement when:
 You want to change computer systems
 You need to upgrade to new products to
stay competitive
 New functions of evolving technology
are introduced
 Database growth
 You need newer, faster, higher density
devices
 Taking advantage of the ever improving
price to performance ratio of new
storage devices
 You just require more flexibility
 You need to relocate you data center
 You need to reduce the foot print of your
storage subsystem within the data
center
 Leveraging data migration to provide
Disaster Recovery solutions
 You want to spread IO evenly across all
the disks in the VG
 You need to align IO access patterns
 Random access
 Sequential access
 You want to protect the data integrity
 Database growth
 Database refreshes
 You need consolidate the space in a VG
or multiple VGs
 You need to troubleshoot an ailing
volume for
 Performance
 Availability (failure boundary)
 You need to separate data into separate
LVs
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
A few LVM migration do’s and don’ts
Do
Don’t
 Consider failure boundaries
 Span multiple frames for temporary
migration purposes
 Add LUNs to the VGs spanning multiple
frames temporarily
 Put sequential accessed LVs on their
own LUN
 Take advantage of the disk subsystems
 Treat Raid array groups as single disks
 Use PP Striping
– Reason: Can dynamically change
the stripe width for PP striping
 Span multiple storage frames in one LV
 Avoid LV Striping
– Reason: Can’t dynamically change
the stripe width for LV striping
 Avoid pladding
– Striping on Striping
 Using the same spindles for data and
logs
 ……………
15
© 2013 IBM Corporation
 ………….
Data and Storage Migration Methods for Power Systems pST541
Data and Storage Migration Methods for
Power Systems
Plan
16
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Planning phase
 A successful data migration always requires substantial
evaluation and planning
 Adequate planning is the critical success factor in a migration
project
 Develop a high level migration plan
 Develop a detailed migration plan
17
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Plan
 Determine whether you can migrate data online or off-line
 Online migrations means data can be moved from Source to Target Platforms with:
– No impact to the end user outside of unscheduled outage windows
– No data I/O loss between the application and the disk storage subsystem
– Very little performance impact affecting the end user
 Off-line migrations means when data is moved:
– Data must be in a known state, typically requiring updates or changes to cease while the
movement occurs.
– Data could be unavailable for an extended periods of time, perhaps several hours or
days.
18
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Planning phase - Example migration methodology plan
Action Item
19
–
Establish a migration management team
–
Gather availability and production schedules
–
Document Change Control procedures and
incorporate into the plan
–
Document the time line for migration activities
–
Announce the migration at least 30 days prior to
the intended target migration date
–
Gather information about the storage server
environment and applications (lists, commands,
scripts and/or drawings)
–
Schedule a pre-migration rehearsal that includes all
the members on the migration team and a data
sampling that will enable the application groups to
appropriately conduct the pre- and post migration
verification process
–
Establish a “Migration Status” call-in process
–
Utilize a “Migration Planning Checklist” to assure
that all of the pre migration planning steps have
been executed
© 2013 IBM Corporation
Assigned
Status
Date
Data and Storage Migration Methods for Power Systems pST541
Establish a migration management team technical migration team
 Team members may include but are not limited to:
– Project manager
– Client (account) manager
– DBA/Application owners
– System administrator
– Network administrator
– Security administrator
– Firewall administrator
– Disk storage administrator
– Backup/Recovery administrator
– SAN fabric administrator
– Hardware CE
– Floor planner
– Cable vendor
– Disaster/Recover administrator
– IT Architect
20
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Gather availability and production schedules
 Some examples of application availability may be but are not limited to the
following list:
– Month/end /quarterly processes
– FlashCopy® or Metro/Global mirror/copy running processes and their time restrictions.
– Database/application refreshes
21
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Planning Phase - Example drawings may look like this:
 For example you may want to go from an Hitachi 9980 on the left to a DS8000/SVC on the
right
 Preparation of physical/zoning cabling setup from the backend through the SAN fabric to the
hosts and SVC
Right Controller(2)
Left Controller(1)
CHA P
1P 1Q
ABCD
CHA P CHA P
CHA P
1R 1S
J KLM
E F GH
N PQR
Current
Layout
CHA P
E F GH
J K LM
50060E80039C6202
N PQR
ABCD
E F GH
CHA P CHA P
CHA P
1P 1Q
2X 2Y
2V 2W
ABCD
Right Controller(2)
Left Controller(1)
1R 1S
J KLM
2V 2W
N PQR
ABCD
E F GH
48
52
53
Frame 1
Brocade 1
46
19
35 62
Port
8
28
46
47
50060E80039C6218
Brocade 2
SAN
42
06
38
21
Port
01 02 03 04 05 06 07 08
01 02 03 04 05 06 07 08
10000000C92D1A97
fcs0
5V-08
fcs2
3p-08
10000000C93D6ADA
10000000C93830A5
Host Server AIX hrinprd
22
N PQR
50060E80039C6212
8
50060E80039C6208
fcs3
3v-08
2X 2Y
J K LM
© 2013 IBM Corporation
Even SAN Fabric
ODD SAN Fabric
09 10 11 12 13 14 15 16
09 10 11 12 13 14 15 16
fcs1
5b-08
P1 P2 P3 P4 P1 P2 P3 P4 P1 P2 P3 P4 P1 P2 P3 P4
10000000C93D72C3
SVC
HBA 1
HBA 2
Node 1
HBA 1
HBA 2
Node 2
HBA 1
HBA 2
Node 3
HBA 1
HBA 2
Node 4
Data and Storage Migration Methods for Power Systems pST541
Planning phase – design requirements
 Understanding the requirements may help simplify migration process
Action
Item
Application Environment
Databases to be moved (DB2, Informix®, Oracle®, SQL, Sybase)
Database version
Database size
Availability requirements of databases (any existing SLA’s, downtime issues
to consider)
Cluster environment (MSCS, Veritas, Sun, HACMP™, MC/Service Guard,
etc.)
Action
Item
Network Environment (if applicable)
Topology
Speed of network
23
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Planning phase – design requirements
 Understanding the requirements may help simplify migration process
Action
Item
Storage Environment
Storage Vendor and model (EMC, HDS, IBM, STK, Dell, HP)
Channel type (ESCON, FICON, Fibre, iSCSI, SAN)
SAN HBA & Model (Qlogic, Emulex, JNI)
Number of Channel Paths
Logical to Physical mapping (i.e. RAID-1 vs. RAID-5)
Number of Source volumes to be migrated
Volume sizes
Identify Target volumes to receive source data
24
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Planning summary - Example migration methodology approach
 Utilize a “Migration Planning Checklist” to assure that all of the pre migration planning steps
have been executed.
Action
Migration and validation methodology checklist
Item
Based on the information gathered in the planning phase, structure the
migration architecture to match the production requirements
Use checklists to ensure any operating patches and software are at the
correct levels
Build detailed migration procedures following the chosen architecture
Put together a schedule of events with time lines to implement the migration
procedures
Establish an initial test plan to validate the initial installation of all required
components
Develop a cooperative deployment plan
Write and configure any automation scripts that will speed up the process
Run a simple initial test plan that validates the migration process
Implement the migration procedures and time line built in the design phase
Verify the migration completion by checking the successful completion and
status of the migration jobs
25
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Data and Storage Migration Methods for
Power Systems
Execute
26
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Execute
 During the migration phase, you will need to:
– Communicate your plans
– Obtain, install and configure any necessary:
• Hardware
• Software
• Automation scripts and tools (to perform the actual data migration)
27
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Execute
 An example migration may go as follows:
–This high level illustration is the execution migratepv –l
28
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Execute summary - Example migration methodology approach
 LVM using migratepv
Command
Process and explanation
lsvpcfg
Identify and assign the DS8000 LUNs to the targeted AIX host server
“
Identify the ESS source and DS8000 targeted LUNs on the host server
“
bootinfo -s
Identify the sizes of the DS8000 target LUNs
extendvg
Move the DS8000 LUNs into the VGs appropriately
lsvg-p
Verify the DS8000 LUNs are added to the VG
lsvg -l
Identify the logical volumes (LVs) to migrate
migratepv -l
lv_name
Copy LV data from the ESS source LUNs to the DS8000 target LUNs
lsvg -p
vg_name
Verify the LUNs are copied
rmdev -dl
Remove the ESS source LUNs from the VGs
rmdev -dl
Delete the device definitions from the host ODM
Lsdev –Cc disk Verify the device definitions are are removed
In the ESS, unassign the LUNs from the host server
Source: If applicable, describe source origin
29
© 2013 IBM Corporation
Storage
Migration
Methods
Data and Storage
Migration
Methods for Power Systems
pST541
Execute summary - Example migration methodology approach
 LVM using mklvcopy
Command
What is does
lsvpcfg
Identify and assign the DS8000 LUNs to the targeted AIX host server
chvolgrp –dev
Identify the ESS source and DS8000 targeted LUNs on the host server
bootinfo -s
Identify the sizes of the DS8000 target LUNs
extendvg
Move the DS8000 LUNs into the VGs appropriately
lsvg -l
Verify the DS8000 LUNs are added to the VG
lsvg -p
Identify the logical volumes (LVs) to migrate
lslv -l lv_name
Determine how the LVs are spread across the vpaths
mklv -y
lvdummy
Reserve free space on each LUN for an even spread of the data across LUNs
mklvcopy
Copy LV data from the ESS source LUNs to the DS8000 target LUNs
lslv-l
Verify the lv copies are made
syncvg
Syncronize the LV data from the ESS source LUNs to the DS8000 target
LUNs
lsvg -l
Verify that the sync isn't showing stale, it should show as sync’d
30
© 2013 IBM Corporation
Storage
Migration
Methods
Data and Storage
Migration
Methods for Power Systems
pST541
Execute summary - Example migration methodology approach
 LVM using mklvcopy (continued)
Command What these commands do
lslv -l
Verify the source and target LUNs for each lv
rmlvcopy
Remove the source copy of the lv from the ESS LUNs
lsvg -p
Verify that all the source ESS LUNs are free with no data
reducevg
Remove the ESS source LUNs from the VGs and verify the removal
rmdev -dl
hdisk#
Delete the device definitions from the host ODM
lsdev -Cc
disk
Verify the device definitions are are removed
In the ESS, unassign the LUNs from the host server
31
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Data and Storage Migration Methods for
Power Systems
Validate
32
© 2013 IBM Corporation
Storage
Migration
Methods
Data and Storage
Migration
Methods for Power Systems
pST541
Validate
 It is important to validate that you have the same data and functionality of
the application after the migration
 You should make sure that the application runs with the new LUNs, that
performance is still adequate, that operations and scripts work with the
new system
33
© 2013 IBM Corporation
Storage
Migration
Methods
Data and Storage
Migration
Methods for Power Systems
pST541
A sample validation list may include but not be limited to the
following items:
 Compile migration statistics
 Prepare a report to highlight:
–What worked
–What didn’t work
–Lessons learned
• Share the report with all members of the migration team
• These types of reports are critical in building a repeatable and consistent process
through continuous process improvement, building on what worked and fixing or
changing what didn’t work. Further, documenting the migration process can help you
train your staff, and simplify or streamline the next migration you do, reducing both
expense and risk
34
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Data and Storage Migration Methods for
Power Systems
Overall summary
35
© 2013 IBM Corporation
Storage
Migration
Methods
Data and Storage
Migration
Methods for Power Systems
pST541
Migration Methodology Summary
Evaluate
–
–
–
–
–
–
–
–
–
36
Analyze business
impact
Risks
Business
interviews
Criticality of data
being moved
Performance
Migration types
Key factors
Multi-vendor
environment
requirements
Application down
time
Plan
–
–
–
–
–
–
–
–
© 2013 IBM Corporation
Determine
migration
requirements
Identify existing
environment
Define future
environment
Create migration
plan
Develop design
requirements
Migration types
Create migration
architecture
Develop test plan
Execute
–
–
–
–
–
–
–
–
Obtain software
tools and licenses
Communicate
deployment plan
Validate HW & SW
requirements
Customize
Migration
procedures
Install & configure
Run pre-validation
test
Perform migration
Verify migration
completion
Validate
–
–
–
–
–
Run Post validation
test
Perform knowledge
transfer
Communicate
Project information
Create report on
migration statistics
Conduct migration
close out meeting
Data and Storage Migration Methods for Power Systems pST541
Data and Storage Migration Methods for
Power Systems
Questions and
Answers
37
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Thank you!
For you interest
and attendance
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Data and Storage Migration Methods for
Power Systems
Backup slides
39
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
A deeper dive
Physical to logical makeup
40
© 2013 IBM Corporation
40
Data and Storage Migration Methods for Power Systems pST541
NO…Saying SAN Volume Controller doesn't count!
What is SVC?
SVC Provides Flexibility
Across Entire Storage Infrastructure
Manage the
storage pool from
a central point
Make changes to
the storage without
disrupting host
applications
Volume
Volume
Volume
Volume
SAN
Apply copy
services across
the storage pool
SAN Volume Controller
Advanced Copy Services
Storage Subsystems
DS8000
15K
rpm
41
© 2013 IBM Corporation
HDS
SATA
DS4000
RAID 5
EMC
RAID 1
HP
JBOD
Combine the
capacity from
multiple arrays on
frames into storage
pools
Data and Storage Migration Methods for Power Systems pST541
SVC - From Physical to Logical View
Mapping to Hosts w/SDD or supported MultiPath Driver
Space-efficient
Volume
Storage
Pool
Stripe
16MB – 2GB
Managed
Disk
LUN
42
vdisk0
125GB
vdisk1
10GB
vdisk2
525GB
mdiskgrp0 [EMC Group]
400GB
vdisk3
1500GB
vdisk4
275GB
vdisk5
5GB
mdiskgrp1 [IBM Group]
600GB
mdisk0
100GB
mdisk1
100GB
mdisk2
100GB
mdisk3
100GB
mdisk4
200GB
mdisk5
200GB
mdisk6
200GB
EMC
100GB
EMC
100GB
EMC
100GB
EMC
100GB
IBM
200GB
IBM
200GB
IBM
200GB
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Examples of correct Host to SVC Cluster zoning
vdisk3
vdisk1
vdisk4
vdisk2
Preferred path for vdisk1 is SVC
N1P2 & N1P3
Preferred path for vdisk2 is SVC
N2P2 & N2P3
Non Preferred path for vdisk1 is
SVC N2P2 &N2P3
Non Preferred path for vdisk2 is
SVC N1P2 &N1P3
vdisk1
43
vdisk2
© 2013 IBM Corporation
vdisk3
vdisk4
Data and Storage Migration Methods for Power Systems pST541
DS8000 Hardware Physical Makeup
Is it important to know the physical makeup?; Does it really matter?
Summary - DS8700 (2-way/4 way base frame)
242x model 941
99,999% = ¼ day in 72 years MTBF
44
© 2013 IBM Corporation
Familiar Layout
Data and Storage Migration Methods for Power Systems pST541
Physical to Logical - peeling back the layers
Even numbered extpools
Primary IO Data flow
ownership
Odd numbered extpools
Primary IO Data flow
ownership
Balance
Just like an onion -virtualization has many layers
A
45
B
Arrays
across
Enclosures
© 2013 IBM Corporation
45
Data and Storage Migration Methods for Power Systems pST541
How does the DS8000 virtual layer work?
Raid-0 only
46
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
DS8000 Virtualization layers (Concepts & Architecture)
How logical extents in ranks are formed from the DS8000, 6+P+S type array format
EXT
EXT 11
1GB
47
EXT
EXT22
1GB
© 2013 IBM Corporation
EXT
EXT
3 3
EXT
EXT33
1GB
EXT 4
EXT
EXT 44
1GB
EXT 5
EXT
EXT55
1GB
Data and Storage Migration Methods for Power Systems pST541
Since the time to complete an I/O operation depends on:
What causes disk latency?
Does Cache mitigate disk latency?
LUN
1
LUN
1
a. Seek time- time to position the read/write head
b. Rotational delay- time waiting for disk to spin to proper starting point
c. Transfer time
RAID-5 6+P
Made up of strips from the outer section/edge of each physical Disk
You could deduce that:
a)
Logical-disk3 would be a better place to store data that will be randomly accessed since the read/write heads would
most
likely
shorter
times
middle
of the
disks.
LUN
3 3have Made
up seek
of strips
from to
thethe
middle
sections
of each
physical Disk
RAID-5 6+P
LUN
b) Logical-disk1 would provide greater sequential throughput since it is on the outer edge of the disks.
EXT
LV 1 1
48
EXT
LV 22
© 2013 IBM Corporation
EXT
EXT
3 3
EXT
LV3 3
EXT 4
EXT
LV4 4
EXT 5
EXT
LV5 5
Data and Storage Migration Methods for Power Systems pST541
LUN Sharing Best practices … is it OK?
 You should know!
LUNs sharing the same array/rank in one extpool
49
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
IBM XIV Storage Distribution Algorithm
 Each volume is spread across all drives
 Data is “cut” into 1MB “partitions” and stored on the disks
 XIV algorithm automatically distributes partitions across all disks in the system
pseudo-randomly
Interface
Interface
Interface
Switching
Data Module Data Module Data Module
50
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
XIV Distribution Algorithm on System Changes
 Data distribution only changes when the system changes
– Equilibrium is kept when new hardware is added
– Equilibrium is kept when old hardware is removed
– Equilibrium is kept after a hardware failure
Data Module 1
Data Module 2
Data Module 3
Data
4
NodeModule
4
[ hardware upgrade ]
51
© 2013 IBM Corporation
Storage
Migration
Methods
Data and Storage
Migration
Methods for Power Systems
pST541
XIV Distribution Algorithm on System Changes
 Data distribution only changes when the system changes
– Equilibrium is kept when new hardware is added
– Equilibrium is kept when old hardware is removed
– Equilibrium is kept after a hardware failure
The fact that distribution is full and
automatic ensures that all spindles join
the effort of data [re-distribution
after
hardware failure ]
configuration change.
Tremendous performance gains are seen
in recovery/optimization times thanks
Data Module 2
Data Module 1
to this fact.
Data Module 3
52
© 2013 IBM Corporation
Data Module 4
Storage
Migration
Methods
Data and Storage
Migration
Methods for Power Systems
pST541
Tips – What are the most common/Important OS I/O Tuning Parameters?
 Device Queue Depth
– Queue Depth can help or hurt performance per LUN
• • Be aware
of Queue
Depth
when planning
system
layout,
adjust only if necessary
Queue Depth
is central
to the following
fundamental
performance
formula:
IO Rate = Number
Commands
* Response
Timedevice
per Command
• • To calculate
- best of
thing
to do is
go to each
“Information Center” URLs listed in
• link slide
For example:
–
IO Rate = 32 Commands per Second / .01 Seconds (10 milliseconds) per Command = 3200 IOPs
• What
are
the default Queue Depths? ___
Some real-world examples:
• rates
OS=Default Queue Depth= Expected IO Rate
 HBA transfer
•
AIX Standalone = 16 per LUN = 1600 IOPs per LUN
– FC adapters
•
AIX VIOS
= 20 per LUN = 2000 IOPs per LUN
 LVM striping• vsAIX
spreading
VIOC
= 3 per LUN = 300 IOPs per LUN
•
Windows
= 32 per Disk = 3200 IOPS per LUN
 Data Placement
– Random versus sequential
– Spreading versus Isolation
Source: Queue Depth content provided by Mark Chitti
53
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Data Placement and Host Vdisk mapping
 Spreading versus Isolation
– Spreading the I/O across MDGs exploits the aggregate throughput offered
by more physical resources working together
– Spreading I/O across the hardware resources will also render more
throughput than isolating the I/O to only a subset of hardware resource
– You may reason that the more hardware resources you can spread across,
the better the throughput
• Don’t spread file systems across multiple frames
Makes it more difficult to manage code upgrades, etc.
 Should you ever isolate data to specific hardware resources?
 Name a circumstance!
• Isolation
– In some cases more isolation on dedicated resources may produce better I/O
throughput by eliminating I/O contention
– Separate FlashCopy – Source and Target LUNs – on isolated spindles
Source: Slide provided by Dan Braden
54
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Data Placement – What causes THRASHING?___
Most commonly
when
workloads peak
thesame
same time
or log filesresult
and data
share physical spindles
Placing
Applications
on atthe
LUNs/Pools
in files
IO contention
For existing applications,
useonstorage
and
tools to understand current
LUN1 made of strips
the outer edge
of server
the DDMs performance
(1s) also could havemonitoring
App A Raid-5 7+P
application workload characteristics such as:
•
•
•
•
•
•
•
•
Read/Write ratio
1
1
1
1
1
1
1
1
2
2
2
2
2
Random/sequential
ratio
2
2
2
5
5
5
5
5
5
5
5
4
4
4
4
4
4
4
4
3
3
3
3 size (blocksize)
3
3 transfer
3
3
Average
Peak workload (I/Os per second for random access, and MB per second for sequential access)
Peak workload periods (time of day, time of month)
made ofrequirements
strips in the middle
of the DDMs (3s) also
couldRemote
have App Mirroring)
B Raid-5 7+P
CopyLUN3
services
(Point-in-Time
Copy,
Host connection utilization and throughput (HBA Host connections)
Remote mirroring link utilization and throughput
Extent pool or 8 Ranks
Strip1
Strip2
Strip3
Strip4
Strip5
55
© 2013 IBM Corporation
Storage
Migration
Methods
Data and Storage
Migration
Methods for Power Systems
pST541
Data Placement - Storage Pools and Striping
– Should you ever stripe with previrtualized volumes?
– We recommend not striping or spreading in SVC, V7000 and XIV Storage Pools
– Avoid LVM spreading with any striped storage pool
– You can use file system striping with DS8000 storage pools
• Across storage pools with a finer granularity stripe
• Within DS8000 storage pools but on separate spindles when volumes are created sequentially
• Striped Pools
No Host Stripe
• Sequential Pools
Host Stripe
S
t
r
i
p
e
Host Stripe
Host Stripe - Raid-0 only
56
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Random IO Data layout
Disk subsystem
1
2
3
4
5
RAID array
LUN or logical disk
PV
Source: Slide provided by Dan Braden
57
© 2013 IBM Corporation
What does random LV creation order, help prevent
____?
1
2
3
4
5
datavg
# mklv lv1 –e x hdisk1 hdisk2 … hdisk5
# mklv lv2 –e x hdisk3 hdisk1 …. hdisk4
…..
Use a random order for the hdisks for each LV
Data and Storage Migration Methods for Power Systems pST541
Sequential IO Data layout
 Does understanding the backend enable good front-end configuration?
 Sequential IO (with no random IOs) best practice:
– Create RAID arrays with data stripes
– RAID 5 arrays of 5 or 9 disks
• RAID 10 arrays of 2, 4, 8, or 16 disks
– Create VGs with one LUN per array
– Create LVs that are spread across all PVs in the VG using a PP or LV strip
size >= a full stripe on the RAID array
– Do application IOs equal to, or a multiple of, a full stripe on the RAID array
– Avoid LV Striping
• Reason: Can’t dynamically change the stripe width for LV striping
– Use PP Striping
• Reason: Can dynamically change the stripe width for PP striping
Source: Slide provided by Dan Braden
58
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Data Layout - OS Spreading versus Striping
•Is there is a difference? What’s the diff?
– Do you know what your volumes are made of?
– Change Vpath to hdisk here.
File system spread
Source: Redbook SG24-6422-00 IBM 800 Performance Monitoring and Tuning Guide
59
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Data Layout Summary
• Data layout affects IO performance more than any tunable IO parameter!
• Good data layout avoids dealing with disk hot spots
– An ongoing management issue and cost
• Data layout must be planned in advance
– Changes are generally painful
• iostat might and filemon can show unbalanced IO
• Best practice: evenly balance IOs across all physical disks unless TIERING
• Random IO best practice:
– Spread IOs evenly across all physical disks unless dedicated resources are needed to
isolate specific performance sensitive data
• For disk subsystems
– Create RAID arrays of equal size and type
– Create VGs with one LUN from every array (not frame)
– Spread all LVs across all PVs in the VG
60
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Documentation – Does it matter? Why?
Track data placement and Host Vdisk mapping
 Disk mapping at a glance
– Mapping becomes important
 Spreading versus isolation
• Spreading
61
© 2013 IBM Corporation
• Isolation
Data and Storage Migration Methods for Power Systems pST541
Documentation – Why it matters
 How do I achieve SVC node to Server Balance?
• Use the SVCQTOOL listed under the tools section of this slide
deck to produce a spread sheet similar to this Or
• Use the script found in the speaker notes of this slide
• Add a column for preferred node to host client
Spread sheet developed by Keith Williams
62
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Are there any automated storage inquiry tools out there that will
help me understand my setup?
 Storage tools
– Gathers information such as, but not limited to:
• LUN layout
• LUN to Host mapping
• Storage Pool maps
• Fabric connectivity
– DS8QTOOL
• Go to the following Website to download the tool:
http://congsa.ibm.com/~dlutz/public/ds8qtool/index.htm
– SVCQTOOL
• Go to the following Website to download the tool:
http://congsa.ibm.com/~dlutz/public/svcqtool/index.htm
63
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
SA - How do I improve disk performance on the Host?
• Reduce the number of IOs
– Bigger caches
• Application, file system, disk subsystem
– Use caches more efficiently
– No file system logging
– No access time updates
• Improve average IO service times
– Better data layout
– Reduce locking for IOs
– Buffer/queue tuning
– Use SSDs or RAM disk
– Faster disks/interfaces, more disks
– Short stroke the disks and use the outer edge
– Smooth the IOs out over time
• Reduce the overhead to handle IOs
64
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Troubleshooting – What’s the most common thing that changes over time?
Pool 1
An ounce of prevention is worth a pound of cure
Pool 2
Rank
Rank
•
Depending on the work load characteristics,
isolating the workload may prove to be more
Vdisk1
beneficial and out perform a larger array.
App A
–
Rank
There are 3 important principles for creating a logical configuration for the Storage Pools to optimize
performance:
Map
Host A
•
Workload isolation
•
Workload resource-sharing
•
Workload spreading
Rank
Some examples of I/O workloads or files/datasets which may have heavy and continuous I/O access patterns are:
•
•
Sequential workloads (especially those with large blocksize transfers)
Rank
Vdisk2
Log files or datasets
Rank
•
Sort/work datasets or files
•
Business Intelligence and Data Mining
•
Disk copies (including Point in Time Copy background copies, remote mirroring target volumes, and tape simulation
on disk)
•
Video/imaging applications
•
•
•
Engineering/scientific applications
Rank
Certain batch workloads
Map
App B
Host B
Data Migration
I always separate Log files from Data files for best performance.
Apps sharing the same physical spindles on traditional arrays
may peak at the same time
65
© 2013 IBM Corporation
Rank
Data and Storage Migration Methods for Power Systems pST541
StorAdmin – How do I improve disk performance?
• Data layout affects IO performance more than any tunable IO parameter
• If a bottleneck is discovered, then some of the things you need to do are:
– Identify the hardware resources the heavy hitting volumes are on
• Identify which D/A pair the rank resides on
• Identify which I/O enclosure the D/A pair resides on
• Identify which host adapters the heavy hitting volumes are using
• Identify which host server the problem volumes reside on
• Identify empty non used volumes on other ranks – storage pools
– Move data off the saturated I/O enclosures to empty volumes residing on less used
ranks/storage pools
– Move data off the heavy hitting volumes to empty volumes residing on less used hardware
resources and perhaps to the another Storage Device
– Balance LUN mapping across
• Backend and host HBAs
• SVC IOgrps
• SVC preferred nodes
– Change Raid type.
66
© 2013 IBM Corporation
Data and Storage Migration Methods for Power Systems pST541
Troubleshooting: What are some Storage Bottlenecks?
• After verifying that the disk subsystem is causing a system bottleneck, a number of solutions
are possible. These solutions include the following:
• Consider using faster disks, SSD will out perform HDD, etc.
• Eventually change the RAID implementation if this is relevant to the server’s I/O workload
characteristics.
– For example, going to RAID-10 if the activity is heavy random writes may show
observable gains.
• Add more arrays/ranks to the Storage pool.
– This will allow you to spread the data across more physical disks and thus improve
performance for both reads and writes.
• Add more RAM
– Adding memory will increase system memory disk cache, which in effect improves disk
response times.
• Finally, if the previous actions do not provide the desired application performance:
– Off-load/migrate - processing to another host system in the network (either users,
applications, or services).
67
© 2013 IBM Corporation
Download