Ds5000 Redbook (Redbook) Note 3/4

Chapter 2 enhanced remote
Processing ing
P423---P498
 
Enhanced remote memory ing (ERM) --- is the previous rvm (Remote Volume
Processing)

▲Metro tracing ing: Synchronous image mode

▲Global copy: asynchronous write mode, asynchronous copying ing without consistency
Group

▲Global writing: asynchronous writing mode, asynchronous processing ing
Consistency group

 
In erm, 2nd logical drives can be in read-only mode.
In volumecopy, the 2nd logical drives are also in read-only mode.

After mirror is suspended, the 2nd subsystems no longer receive any I/O from the Master System, which is called waiting ing repository.
Volume

After mirror continues, the changed data will be synchronized again instead of all

The switching ing mode can be changed at any time, which is called dynamic mode switching.

 

12.1.1 requirements
Erm requires a separate license, which must be installed on each ds5000 (primary site and secondary
License must be installed on the site)
Ds4200/4700 supports 64 mirror
Pairs
(This is the sum of primary and secondary.
P472)
Ds4800 supports 128 mirror
Pairs
(This is the sum of primary and secondary.
P472)
Ds5000 supports 128 mirror
Pairs
(This is the sum of primary and secondary.
P430)
 
To use erm, San fiber networks must be forced: fiber switches or FC-IP Routers
P426
The optical fiber connection in ERM is dedicated to copying data and is not used for other I/O

12.1.2 erm terminology
Primary Site: Also called local site
Secondary site: Also called remote site
Primary Storage System: Also called local sotrage Subsystem
Second storage subsystem: Also called remote storage subsystem

Primary and Secondary
The models do not have to be the same
P428

Processing storage controller pair
A to
B to B

 

Mirror fiber re Channel Connection
Dedicated to the I/O connection between the primary controller and the secondary Controller
Cannot be directly connected
Because in ERM, San must be used to connect
 
Maximum mirror in ERM
Pairs refers to the sum of primary and secondary.
P430

Only the standard logical drive can join the image relationship.
Secondary logical drive is larger than or equal to primary logical drive
Any logical drive can belong to only one image relationship

 

Controller A in primary must be mapped to Controller A in secondary.
B to B

 

When you delete primary logical drive, mirror relationship is deleted first, and then primary is deleted.
Logical drive
When you delete the secondary logical drive, you must first Delete the mirror relationship and then the secondary
Logical drive

 

Primary logical drive and secondary logical
Drive capacity can be dynamically increased (DVE)
Secondary logical drive must be larger than primary logical drive
Mirror repository drive cannot be dynamically increased

 

Primary and secondary logical drive can be dynamically expanded (DCE)-that is, the hard disk is added to the array.
A maximum of two hard disks can be added at a time.

 

Dynamically changing the segment size in ERM --- (DSS)

 

The raid level cannot be dynamically changed in ERM.

 

Flashcopy can be used in both primary and secondary logical drive

 

Primary logical drive can be the source or target Logical Device of volumecopy
However, secondary logical drive cannot be the source or target Logical Device of volumecopy.

 

Mirror relationship
Including: Mirror repository drive and tracing ing Pairs
Before establishing mirror relationship:
▲Establish appropriate FC or Ethernet communication between storage devices
▲Enable ERM is required for both storage
▲Create ERM on both storage --- create mirror repository drive
▲Create a secondary logical equal to or greater than the primary logical drive on secondary
Drive

12.4.1 remote mirror status
P435
Status of various icons
Finished: Green
Synchronization in progress: There is a clock icon
Not synchronized: there is a red slash
Pause: There is a yellow lightning

 

---------------------------------------------------------------------------------

Regardless of the mirror mode, when a new mirror is created, a full synchronization

Data in the secondary logical drive can only be changed by the processing process

Mirroring mode can be changed at any time --- DMS

 

Only the primary controller records logs in mirror repository logical drive.
Mirror repository logical drive only records logs and does not store data

 

When data needs to be read, it is only read in primary and will not communicate with secondary

 

Mirror repositor logical drive supports a maximum of 128 I/O queues
The status is synchronized before the maximum value is reached.
If the maximum value is reached, the state is unsynchronized, and the host can still write, but the write on the remote site will no longer generate

---------------------------------------------------------------------------------

 

12.5.1 Metro processing ing (synchronous processing)
Synchronous image mode

When primary receives a write request, it first writes the logs in the mirror repository logical drive
At the same time, write the primary logical drive
The controller starts a remote write operation in secondary logical drive.
After the remote write operation is complete, the primary controller deletes the log.
Eventually, the controller sends an I/O signal to the host

 

12.5.2 global copy (asynchronous copying ing without write
Consistency group)
Asynchronous write mode
When there is a write primary, write immediately and report completion
It is not guaranteed that the write order on the primary site is the same as that on the remote site.

 
Write Request
---> Controller logs in mirror repository logical drive
---> At the same time, write data
---> Returns an I/O completion signal.
---> Write data to secondary site in the background
---> Primary Delete log
 
8.5.3 global processing ing (asynchronous processing ing with write
Consistency group)
Asynchronous write mode
It will protect the write order in secondary logical drive: tracks the write order, queues it and sends it to the remote controller in the same order

 

Write Data
---> Primary in mirror repository logical
Logs in Drive
---> Write data
---> Send a completion signal to the host
---> Copy the affected data blocks to secondary (remote site receives the same write order)
---> Delete logs

 

12.5.4 data resynchronization Process
When the remote site link is interrupted or the logical device fails
Primary changes to the unsynchronized status and sends an I/O completion signal to the host.
The host can still be written in primary logical drive, but "write" won't happen again on secondary"
Write requests will be stored in the primary remote Repository

After the connection is restored, a resynchronization is generated.

 

When the status is unsynchronized/suincluded, the host can write the primary

 

Unsynchronized and sucommitted ded are essentially different. An error occurs in both states,
When suchronized is enabled, you must manually change the status to synchronized.
When unsynchronized is enabled, the state can be automatically or manually changed to synchronized.

 

When a new communication is established, only changed data blocks are sent to the remote site.

 

Full synchronization is created only in the following cases:
▲New mirror relationship
▲All members in the image relationship are failure
▲Mirror repository logical drive failure
▲After a communication error occurs, all data block tracking is changed in mirror repository logical drive.
▲After suincluded status, all data block tracking is changed in mirror repository logical drive.

 

Manual resynchronization: recommended method
Automatic resynchronization: automatically starts Synchronization After the communication is detected to be restored.

 
12.5.5 Data Synchronization priority
A total of five levels. This parameter is only useful when you re-sync
After changing the level, it takes effect in both primary and secondary.
 
12.6 San fabric and Ethernet connectivity
Must contain San ports, San zoning, and cabling considerations

12.6.1 San fabric and San zoning Configuration
San fabric configuration: dedicated Remote Communication ing ports (A2 and B2 host
Side controller ports) must be attached
A San fabric with support for the Directory Service and name
Service interfaces.
There must be a dedicated zoning on the SAN Switch.

 

More than two San switches are recommended for High Availability redundancy.

San fabric zoning: mandatory San zoning.

Do not use a hybrid zoning definition on a SAN Switch (for example, use both port and wwpn)

Use the outermost host port in ERM
P444

 
Ds4000 storage server Fiber Channel configuration for erm
Erm requires two dedicated controller host ports to connect to each storage system.

▲Use the outermost host port in ds4200/4700/4800 as the mirror connections
This pair of ports will not receive host I/O requests and will only be used for communication with another storage.

 
Fiber Channel distance limits: distance between two switches
Use 9um Single-Mode Optical Fiber for 10 km

--- The distance will conflict with the speed. When the maximum distance is reached, the speed will be automatically reduced (4G is reduced to 2G, 2g is reduced to 1G );

--- At full speed 4G, only m distance is supported
50um multimode optical fiber is 0.5 km

 

The global copy and global mirror modes support over 5000 km.

 

12.6.2 Ethernet management network configuration for erm
Force the management workstation (smclient) to access both sites over Ethernet (primary and secondary)
Therefore, the two storage subsystems must be set to a subnet or through a VPN
(Ds4000 p491)

 

8.7 enhanced remote tracing ing: Step-by-Step
Erm key must be purchased. This key supports all partitions.

Storage subsystem → premium features → list

Two gray hard disk icons when the status is enable/deactivated

 
Activate erm: storage subsystem → remote processing →
Activate

Then the ERM icon in the lower left corner turns green.

 
Cannot create REM on raid0
P449
Mirror repository logical drive cannot be created on raid0

Two miorror repository logical drivee

 

8.7.2 creating enhanced remote configuring relationships
You cannot use the mirrored lun as the primary logical drive.

 

Select a logical drive ---> logical drive → remote
Processing ing → create
Or right-click ---> Create remote mirror

 

Viewing enhanced remote configuring properties and status
Select a lun ---> logical drive → properties or right-click →
Properties
You can view:

Synchronization progress, image relationship (primary or secondary)

Image status-synchronization completed, in progress, paused, not synchronized
Synchronization priority-lowest, low, medium, high, highest

 

In the profile file, you can see the relevant information about erm.
The number of REM can be seen in the summary.
You can see mirror in logcial drives-> Repositories
Repository Information
The REM attribute is displayed in logical drives-> mirrors.

 

View associated components view mirror relationship
Select primary logical drive-> View associated
Elements

 

12.7.4 changing mirror synchronization settings
The attribute can only be changed on primary and cannot be changed on secondary.

 

Select primary logical drive-> logical drive → remote
Refreshing ing → change → Synchronization
Or right-click Change → synchronization settings

 

Using "automatic re-synchronization" is not the best habit, because data will become inconsistent when it becomes invalid.
P465

 

12.7.5 changing mirror write mode
Select primary logical drive-> logical drive → remote
Refreshing ing → change → write mode
Or right-click Change → write mode

 

8.7.8 suspending and resuming a mirror relationship
When the synchronization relationship is paused, primary's I/O will be recorded to mirror repository logical.
Drive and data is written to primary logical drive

When the asynchronous relationship is suspended, the primary controller simply stops the Remote Write Request.

 

Select primary logical drive-> logical drive → remote
Processing → suspend
Or right-click Change → suspend processing ing

 

When paused, the icon turns gray with a lightning icon

 

Resume restore ing: no special icon after recovery

 

12.7.9 removing mirror relationships
Deleting the image relationship does not delete primary logical drive, secondary logical drive, and mirror.
Repository logical drive;
Primary logical drive is still accessible

Delete is only valid for local logical drive
First, delete the local logical drive, and then delete the remote logical drive. If you cannot delete the remote
Logical drive, the communication is faulty, and an error is reported.

Logical drive → remote tracing ing → remove mirror
Relationship

 

8.8 Erm and disaster recovery
In the event of a disaster, Primary and Secondary logical drives can be converted to each other.

 

Changing a secondary logical drive to a primary logical
Drive
Select secondary logical drive, logical drive → remote tracing ing → change
→ Role to primary

 

Changing a primary to a secondary logical drive
Select primary logical drive ---> logical drive → remote
Grouping → change → role
Secondary

 

Forcing a secondary to change to a primary logical drive

Forcing a primary to become a secondary logical drive
If a communication error occurs when secondary is converted to primary, an error message is displayed.

 

View the recovery guru when an error occurs.

 

12.8.2 re-establishing Remote Authentication ing after failure
Recovery

 

12.9 enhanced remote processing ing solution design
Psung
▲Minimum redundancy is used for intra-site configuration: Two FC switches are connected to two ds4000
▲Redundant optical fiber is used for site configuration: Two FC switches are connected to two ds4000, and ISL (Inter switch) is used between switches.
Link) Connection
▲High availability sites are configured with redundant Optical Fiber: Four FC switches are connected to two ds4000 full redundancy structures, and ISL (Inter switch) is used between switches.
Link) Connection

When the speed is 4 GB, the distance between primary and secondary is 150 m

12.10 erm Maintenance
When mirror relationship becomes invalid, the mirror State becomes unsynchronized.
State (with a red slash)
Check the recovery guru

 

In synchronization mode, the FC link is interrupted: it becomes the unsynchronized mirror state, the host is still writable, and remote data is not written.
FC link interruption in asynchronous mode: Primary periodically tries to connect to secondary; In mirror repository logical
Drive stores the written sequential logs. The status is synchronized before the mirror repository is full. When mirror
After the repository capacity is fully written, the status changes to unsynchronized, the host is still writable, and remote data is not written.

 

12.10.3 fibre channel mirror link test function
Test mirror communication: The test passes a green light and the test fails a red light.

 

12.10.4 secondary logical drive Error
When secondary logical drive is incorrect, primary is also identified as unsynchronized

 

12.10.5 primary controller failure
When the previous controller of primary becomes invalid, All I/O will be automatically transferred to another controller.

 

12.10.6 primary controller Reset
If the data can be normally written to secondary, the preferred Controller on primary cannot be changed.

 

12.10.7 secondary controller failure
When the secondary controller becomes invalid, primary will not communicate with secondary, and the status will change to unsynchronized.

 

12.10.8 write consistency group and unsynchronized state

 

12.11.1 synchronization priority
When the highest priority is synchronized for 8 times, the lowest is synchronized for 1 time.
When the highest priority is 6
During secondary synchronization, low
1 time
When the highest priority is 3.5 times for synchronization, medium is 1 time
When the highest priority is synchronized twice, the lowest is synchronized once.

 

12.11.2 synchronization performance and logical drive
Settings
▲Set the same write cache on both buckets
▲Do not use read cache on secondary
▲Create an array
▲Set up raid1 on secondary for Load Balancing

 

12.11.3 processing ing mode and Performance
-Metro processing ing is a "must synchronize" mode.
-Global copy and global mirror are a "can synchronize" modes.
-Global mirror needs more storage controller processing resources
To enforce the host I/O Write Request order.

 

12.11.4 sorting ing connection distance and Performance
When the distance exceeds 10 km, you can only use the asynchronous image mode.
When the distance is too long, the system will automatically slow down to avoid conflict.
When the distance is too long, you can use an IP address to connect

 

Asynchronous mode is used when the load is too heavy

 
12.12 long-distance ERM
Use global copy and global mirror operating
Modes, using an asynchronous image, with a distance of more than 5000 km
 
Bytes --------------------------------------------------------------------------------------------

Some questions in ds4000:
How can I/O on mirror repository be determined ???
P488 Figure 8-11 is there a problem ???
P524 figure master/secondary swap in remote Images
Is there a primary/secondary behind the subsystem ??
P531 8.8.2 re-establishing
Remote troubleshooting after failure
Rediscovery requires lab
Bytes ---------------------------------------------------------------------------------------------

This chapter is similar to Chapter 8th of ds4000.
Bytes ---------------------------------------------------------------------------------------------

PM
 
 
 
Chapter 4 ds4000 Performance
Tuning
P499---P528

13.1 workload types
▲Transaction-based: high number of transactions (iops), in IPOs
▲Throughput-based: large I/OS, in Mbps

Workload = [transactions (number of host iops)] * [Throughput
(Amount of data sent in one Io)]

The transactions and throughput in the workload are mutually exclusive. The transactions increase and the throughput decreases. vice versa;

Transaction based processes (iops): In iops
Throughput Based processes (mbps): In Mbps

There are different parameter settings for transaction-based and throughput-based workloads.

 

13.2 solution-wide considerations for performance
1. Consider the location and path of the device/logical device when the host accesses the Lun.
2. Consider the location of data in the storage array and logical device
3. Consider the array where the number is located in the backend device and how the data passes through the Members in the array, such as the number, size, speed, and application conflict of devices.
4. Configuration conflict: consider how the host uses Storage

13.3 host considerations
13.3.1 host based settings
Some hosts can change the value of write cache and cache read-ahead, which affects both transaction and throughput workloads.

▲Block size: used to determine the segment size. It is recommended that when a high transaction workload occurs, the segment size should be at least 2 times that of I/O.
Blocksize
▲Queue depth: the queue depth on ds4000 cannot be exceeded (the queue depth on ds4000 is limited by firmware and firmware
6.1 is 2048)

All Lun shared queue restrictions on Storage
Formula: queue depth on ds4000/
(Number of hosts * Luns on each host ;)
Maximum number of Luns selected on the host
For example, if four hosts have 2048, 14, 16, and 32 Luns, the queue depth is/(4*32) = 16

Some hosts define the queue depth on the hosts card
Calculation formula: Number of queue depth/memory cards on ds4000

The queue depth is too large, which may cause data loss and file corruption.

When FC tape and FC disk are used, they must be separated on different HbA cards. Therefore, the data styles of these two devices vary greatly.

Host Data Layout
▲In raid1/10, the Lun on the same Array uses the preferred path through different controllers.
▲In RAID5, it is best to place the Lun on a different array (that is, a lun on an array) and use the preferred path through different controllers.
▲In raid0, the stripe size must be completed the size of the underlying Stipe
Width defined for the logical drive

?????

 

13.3.2 host setting examples
AIX operating system settings
Transaction settings: AIx can no longer modify the cache read-ahead
Use the chdev command to change the attribute values.
# Chdev-l hdisk #-A queue_depth = xxx
-P: Change the disk's queue depth.
# Chdev-l FCS #-A num_assist_elem = 256
-P
Change the num_assist_elem of the HBA ---- 256 is recommended for High Performance
P157

Throughput Based settings
Set throughput in Aix and lg_term_dma and max_xfer_size on the memory card

Windows operating system settings
Fabric settings: Set queue depth in the queue card
Disk types: Basic disk (default) or Dynamic Disk

In Windows, you can use the Extend Command of diskpart.exe to perform basic partitioning on the disk.

Basic disks do not support Soft Raid.

Extend of diskpart.exe can be used in windows.
Command to base the primary partition on the disk
P159---P161

Extend [size = N] [disk = N] noerr

C:/> diskpart.exe
Diskpart> List volume
Diskpart> select Volume 3
Diskpart> extend [size = N] [disk = N]
Noerr
If no size is specified, no unallocated space is extended. If no disk is specified

The Extend Command can only be used for NTFS partitions.
I/O does not need to be stopped during expansion

 

13.4 application considerations
When you use a mix of transaction and throughput-based workloads, consider transaction performance first.

Transaction environments: the database, mail, web, and file server are mostly based on transactions.
Throughput environments: fewer transactions, but larger I/O

 

13.5 ds4000 storage server considerations
13.5.1 which model fits best
First, consider selecting a suitable storage model

13.5.2 Storage Server Processes
Ds4000 copy services

Enhanced remote tracing ing (ERM ):
When the highest priority is synchronized for 8 times, the lowest is synchronized for 1 time.
When the highest priority is 6
During secondary synchronization, low
1 time
When the highest priority is 3.5 times for synchronization, medium is 1 time
When the highest priority is synchronized twice, the lowest is synchronized once.

Flashcopy
Function: When flashcopy is no longer required, it should be stopped because the source Lun cannot be written during copy-on-write. Although the time is short, conflicts may occur;

After disable, you can re-create
P170

13.5.3 storage server modification functions
Mutually Exclusive between priority and system performance: high priority, affecting performance, but time is short; low priority, high performance, but long time

Media scan: checks all Luns in the background. The default value is once every 30 days.
Defragmenting an array: Used to release space. once started, it cannot be stopped and cannot be changed during running.
Copyback: used to copy data from a hot backup disk to a replaced hard disk. When an invalid hard disk is replaced, copyback automatically occurs from the hot backup disk to the replaced hard disk.

Initialization: Also called formatting. Delete all data.

Dynamic segment sizing (DSS): used on the Lun
Dynamic reconstruction rate
(DRR): Used in Raid 1/3/5 to recreate data on a replacement hard disk or hot spare disk.
Dynamic raid level migration (DRM)
Dynamic capacity expansion (DCE)
Dynamic logical drive expansion (DVE)

 

13.5.4 storage server Parameters
Cache blocksize Selection: Global parameter;

4 kb is recommended for transactions (such as databases;

Throughput-based use of 16 KB

When mixing, decide which type to use more
Cache flush control settings: 80% --- 20%

 

13.5.5 disk drive types
Hard Disk Type (FC or SATA)
Speed
Capacity

SATA hard drive does not support queue
Depth
(Ds4000 best practices p54)
FC hard disk queue depth = 16

Arrays and logical drives
Raid array types
Number of disks per array
Array and logical drive creation: We recommend that you create only one Lun on an array.
Logical drive segments: Use 64-128 K in transaction-based

In Throughput-based, set it to the host I/O size or multiple
Logical drive cache settings: Enable read/write cache for Lun settings to improve performance

13.6 write operations and Storage Manager handling of Arrays
13.6.1 write disk operation
13.7 fabric considerations
 
----------------------------------------------------------------

This chapter is the same as Chapter 4th of ds4000 Best Practices
----------------------------------------------------------------

 
 
 
Chapter 4 ds4000 tuning with typical
Applications
P529---P548

14.1 DB2 database
Use a separate array and separate lun to store database files and Database Log Files

Use riad10 to increase the write rate by 25---30% than RAID5
If RAID5 is used, 8 + 1 parity disk is used, and the segment size is 64kb or 128kb.

DB2 logs and archive files: Large workloads and continuous generation. We recommend that you place the base on raid 10.

 

14.2 Oracle databases
Use riad10 to increase the write rate by 25---30% than RAID5
If RAID5 is used, 8 + 1 parity disk is used, and the segment size is 64kb or 128kb.

Use expansion cabinet Loss Protection
Use a 15 k rpm disk, which is 20-30% higher than the performance of a 10 k RPM Disk
Make sure you have enough hot spare disks.

 

Performance monitoring: Monitoring CPU, memory, disk, Network
Use Performance Monitor in Windows
Run the following command in Linux/Unix:
CPU: SAR, vmstat, mpstat, iostat
Memory: SAR, vmstat
Disk: SAR, iostat
Network: netstat

 
 
14.3 Microsoft SQL Server
Use NTFS file system
Use raid1/10
Use a 15 k rpm Disk
For the best performance, 5-12 hard disks are used in the array

 

14.4 IBM Tivoli Storage Manager Backup Server
In TSM, cache is set to 16 KB

 

14.5 Microsoft Exchange 2007

 

----------------------------------------------
Summary:
Use raid1/10
Use high-speed hard drive
Use expansion cabinet Protection
Make sure there is enough hot spare Disk
Enable read/write Cache
Maintain the appropriate number of hard disks in raid
----------------------------------------------
This chapter is similar to ds4000 best practices.
----------------------------------------------
PM