Huawei HyperClone Technical White Paper PDF

Doc.
code

HyperClone
Technical White Paper

Issue 01
Date 2012-04-20

HUAWEI TECHNOLOGIES CO., LTD.

HyperClone Technical White Paper CONFIDENTIAL

2013-7-29
Copyright Huawei Technologies Co.,
Ltd. 2012. All rights reserved
Page 2 of 14

Copyright Huawei Technologies Co., Ltd. 2012. All rights reserved.
No part of this document may be reproduced or transmitted in any form or by any means
without prior written consent of Huawei Technologies Co., Ltd.

Trademarks and Permissions
and other Huawei trademarks are trademarks of Huawei Technologies Co., Ltd.
All other trademarks and trade names mentioned in this document are the property of their
respective holders.

Notice
The purchased products, services and features are stipulated by the contract made between
Huawei and the customer. All or part of the products, services and features described in this
document may not be within the purchase scope or the usage scope. Unless otherwise
specified in the contract, all statements, information, and recommendations in this document
are provided "AS IS" without warranties, guarantees or representations of any kind, either
express or implied.
The information in this document is subject to change without notice. Every effort has been
made in the preparation of this document to ensure accuracy of the contents, but all
statements, information, and recommendations in this document do not constitute the
warranty of any kind, express or implied.

Huawei Technologies Co., Ltd.
Address: Huawei Industrial Base
Bantian, Longgang
Shenzhen 518129
People's Republic of China
Website: http://www.huawei.com
Email: support@huawei.com


2013-7-29
Page 3 of 14

Chapter 1 Overview
As services of enterprises become more diverse and requirements of applications
on centralized storage become stricter, customers put forward higher requests for
online data utilization and protection. The snapshot technology is one of the
effective methods to prevent online storage devices from losing data.
The Storage Network Industry Association (SNIA) stipulates that the snapshot is a
fully usable copy of a defined collection of data that includes an image of the
source data as it appeared at the point in time at which the copy was initiated. A
snapshot may be either a duplicate or a replicate of the data it represents.
The snapshot can be realized through multiple methods. The virtual snapshot and
split mirror (that are called HyperImage and HyperClone in Huawei storage
products) are two most commonly used technologies in the storage industry. The
Huawei OceanStor storage array supports both the HyperImage and HyperClone
to provide customers with multiple choices for online data protection. This
document only describes the HyperClone technology.
Take a data backup that is indispensable to data protection for example.
Traditional backup tasks have to interrupt services for obtaining complete and
consistent backup data. However, any service interruption means commercial
loss in the cut-throat competition. Except the data backup, multiple scenarios
need to use data concurrently, such as report generating, testing, data mining,
and decision support. How can data be used concurrently and how to protect data
in time? The HyperClone answers these questions. The HyperClone creates a full
physical data copy at a certain time, by which to protect data in the system.
Customers can inflexibly create multiple consistent copies at different time.
Chapter 2 The HyperClone of the OceanStor
Storage Array
The split mirror of the OceanStor storage array is named HyperClone.
2.1 Working Principles of the HyperClone
In general, the source volume and copy volume are named the master volume
and slave volume respectively. For the HyperClone, they are named the master


2013-7-29
Page 4 of 14

LUN and slave LUN. Customers can obtain a full physical copy of the master LUN
at a certain time through synchronizing and splitting the master and slave LUN
without service interruption. Any read or write operation to the physical copy does
not impact the master LUN data. Therefore, through the HyperClone, data can be
concurrently used for online backup, data mining, and application testing. The
HyperClone is realized through combining bitmap and copy-on-write, or bitmap
and double write that means to write data onto the slave LUN before onto the
master LUN. The principles to realize the HyperClone are as follows:
If you create a HyperClone group without adding the slave LUN, the principle
of reading or writing the master LUN of the HyperClone group is the same as
that of reading and writing the LUN without the HyperClone feature.
After adding a slave LUN to the HyperClone group, data in the master LUN
will be automatically synchronized to the slave LUN. The copy progress is
displayed until the copy is complete.
If the master LUN receives a write request from the host when performing
initial synchronization (see Figure 1), the synchronization progress needs to
be checked:
If the data block to which new data to be written is not yet copied to the
slave LUN (see the situation ), report success to the host after the
data is written onto the master LUN and then the data block can be
synchronized to the slave LUN.
If the data block to which new data to be written has been copied to the
slave LUN (see the situation ), the data needs to be written onto both
the master and slave LUN.
If the data block to which new data to be written is being copied to the
slave LUN (see the situation ), the data needs to be written onto both
the master and slave LUN after the replication is complete.


2013-7-29
Page 5 of 14

Figure 1 A write request from the host when performing initial synchronization

A write request from
the production host
Primary LUN
When the data block to be written is being synchronized, to perform double
write after the synchronization is complete
When the data block to be written has bee synchronized, directly to perform
double write.
When the data block to be written is not synchronized yet : directly write data
to the master LUN
Initial synchronization
Seconday LUN
Progress
bitmap
Initial synchronization
Data block d in synchronization
Data block q in synchronization
After the host writes data
0 indicates the data
block bitmap that has
been copied; 1
indicates the data
block that has not
been copied, or is
being copied.

Data on the master LUN is the same as that on the slave LUN after the initial
synchronization is complete. If the master LUN receives a write request from
the host at this time, refer to the situation in Figure 1 to write data to the
master and slave LUN.
The master and slave LUN can be split when the initial synchronization is
complete. Both of them can be independently used for data analysis and
testing. They do not interact due to their data change that is recorded in the
progress bitmap, see Figure 2.
Figure 2 A write request from the host after HyperClone splitting


2013-7-29
Page 6 of 14

A write request
from the host
Complete division after initial
synchronization
Progress bitmap
Master
LUN
Receive a write request
from host during division
Slave
LUN
Record
bitmap
Record
bitmap

Data change can occur to both the master and slave LUN after they split.
You can start an incremental synchronization process according to the
differences recorded in the progress bitmap to back up the data of the
master LUN. If the host sends write data requests during the incremental
synchronization, its principle is the same as that of initial synchronization.
If you need to backup the slave LUN data, you can also create a data copy
from the slave LUN to the master LUN according to the differences recorded
in the progress bitmap, which is called incremental reverse synchronization.
During incremental reverse synchronization, the slave LUN protection mode
and non slave LUN protection mode (see Figure 3 and Figure 4 respectively)
need to be distinguished when the host sends a write request. In slave LUN
protection mode, the slave LUN data does not change due to the master
LUN data change during the incremental reverse synchronization. In both
modes, a prior synchronization operation, that is called copy-on-write, will be
performed to a data block to be written before it is synchronized. If the data
block to be written is being synchronized, wait until the synchronization is
complete. The differences between two modes mainly lie in the following two
aspects:


2013-7-29
Page 7 of 14

I. In slave LUN protection mode, data can be directly written to the master LUN
after a certain data block is synchronized and the differences are recorded in the
bitmap. In non slave LUN protection mode, data will be double written onto both
the master LUN and slave LUN.
2. When all synchronizations are complete, slave LUN protection mode requires
splitting the master and slave LUN. At the same time, the difference bitmap and
progress bitmap need to be exchanged.
Figure 3 A write request from the host during incremental reverse synchronization
slave LUN protection mode

Figure 4 A write request from the host during incremental reverse synchronization
slave LUN protection mode


2013-7-29
Page 8 of 14

A write request from
the host
When the data block to be written is being synchronized, write data
to the mater LUN after synchronization is complete.
When the data block to be written is being synchronized, directly to
perform double write
When the data block to be written is being synchronized, perform
double write after completing priority synchronization.
Reverse synchronization in progress
Data x in copy
Generate a priority copy to data z
Master
LUN
Slave
LUN
After the host
write data
Progress bitmap
Data R in synchronization

Precautions of using the HyperClone:
1. The capacity of the master LUN must be the same as that of the slave LUN.
2. The master and slave LUN can work only on the same controller. The
working controller and owing controller of the slave LUN is changing with that
of the master LUN.
3. Only when the slave LUN completes synchronization and is split, it can
respond to host services.
4. The simultaneous reverse synchronization task can only be one in a virtual
snapshot group.
2.2 Features of the HyperClone
2.2.1 Supporting One Master LUN plus Eight Slave LUNs Mode
The HyperClone of the OceanStor storage array supports one master LUN plus
eight slave LUN mode (see Figure 5). In this mode, eight progress bitmaps record


2013-7-29
Page 9 of 14

the differences between eight master LUNs and eight slave LUNs. The data
written to the master LUN will be written to eight slave LUNs accordingly. The
HyperClone in this mode can back up eight sets of source data, which will be
applied into different data analyses.
Figure 5 One master LUN plus eight slave LUNs mode

Progress bitmap 1
Progress bitmap 8
Master LUN
Slave LUN
Slave LUN 8

2.2.2 Zero Backup Window
The traditional backup often leads to poor host performance and even
unacceptable service. To obtain intact and consistent data, the traditional backup
can be performed only when the application is stopped or the service I/O is small.
The backup window means the maximum backup time the application can
tolerate, in other words, the tolerable downtime of the application. Customers do
not need to stop their devices when performing backup through the HyperClone.


2013-7-29
Page 10 of 14

2.2.3 Uninterrupted Services
During synchronization and reverse synchronization process, the HyperClone of
the OceanStor storage array can be performed online without services
interruption, which meets customer' demand for online data backup and realizes
that data backup is independent of services.
2.2.4 Supporting Dynamic Copy-Speed Control
When the HyperClone is performed, the performance of the storage array will be
inevitably impacted. The OceanStor storage array supports dynamic copy-speed
control to avoid the conflict between copies and production services. When the
system is detected very busy by the storage array, the copy speed can be
manually slowed down to spare system resources for services operation. When
the system is idle, the copy speed will be dynamically increased.
2.2.5 Supporting Reverse Synchronization
The traditional offline backup data cannot be directly read online. It takes long to
obtain the usable copy at the backup time to realize data recovery. The
HyperClone of the OceanStor supports reverse synchronization mode. When the
master LUN data needs to be recovered due to incompleteness or damage, it can
be realized through performing incremental reverse synchronization from the
slave LUN to the master LUN. Besides, the redirection read function of reverse
synchronization can ensure that data in the master and slave LUN can be read or
written consistently during reverse synchronization. Fast incremental
synchronization can update concurrent services data in time as well as saving
customers' management time.
2.2.6 Supporting Automatic Recovery After Disconnection
When faults occur, the HyperClone will enter disconnected mode. The faults
include that the master LUN and slave LUN are invalid. The method to deal with
I/O when the HyperClone stays in disconnected mode is similar with when the
master LUN and slave LUN are split, which is to write data only to the master LUN
and then record differences. However, if the master LUN fails, it cannot receive
any I/O request from the production host. When faults are removed, the
HyperClone will start to recover data according to the recovery policy. If the
recovery policy is automatic recovery, the HyperClone will automatically enter
synchronization mode to synchronize incremental data to the slave LUN; if the
recovery policy is manual recovery, the HyperClone will enter to be recovered
mode and wait for manual synchronization. When the HyperClone is
disconnected, to apply incremental synchronization will greatly shorten recovery
time after failure and disasters.


2013-7-29
Page 11 of 14

2.2.7 Supporting Multiple Slave LUNs That Correspond to Different
Master LUNs to split Simultaneously
In the OLTP, associated data scattered in different LUNs can be kept consistent at
the same split time only when consistent copies of multiple master LUNs data are
created at the same split time. Take the Oracle database for example.
Management data, service data, and log information are usually allocated in
different master LUNs. When copying data, in order to recover their data to the
same time, you need to create consistent copies for their master LUNs at the
same split time. If their data is recovered to the different time, they will lose their
association with each other, which leads to meaningless data recovery. The
HyperClone of the OceanStor resolves this problem through dividing multiple
slave LUNs. It freezes multiple master LUNs data at the split time to obtain
consistent copies of the master LUN at the same split time.
2.2.8 Supporting Slave LUN Protection Mode
The HyperClone of the OceanStor storage array provides the slave protection
function. When the HyperClone performs reverse synchronization, you can
neither stop host services nor change the slave LUN data.
2.3 Application Scenarios of the HyperClone
2.3.1 Data Backup
The HyperClone of the OceaStor allows customers to operate multiple
simultaneous services to access data through multiple physical copies of the
master LUN (see Figure 6). The simultaneous access will not impact each other
but can satisfy database's application requests of creating physical copies of
multiple LUNs at the same time. The virtual snapshot differs from the HyperClone.
It stores the differential data between different times. That is, what the virtual
snapshot stores is not the intact copy of the source data.


2013-7-29
Page 12 of 14

Figure 6 Concurrent services of the HyperClone

Backup server
Testing server
Data analysis
Upgrade drill
Slave
LUN
Slave
LUN
Slave
LUN
Slave
LUN
Master
LUN

2.3.2 Data Recovery and Protection
The HyperClone of the OceanStor can protect source services data. When the
master LUN data suffers virus attacks, and human or physical damage, data can
be recovered through copying the slave LUN to the master LUN in reverse
synchronization mode after choosing data copies at a proper time. When
performing data recovery, you do not need to stop host services in order to
ensure continuous services. Compared with the HyperClone, the snapshot does
not store source data copies. Once source data fails, it is hard to recover data.
2.3.3 Backup Software, VSS, and HyperClone
After Windows 2003 SP1 was released, Microsoft launched Volume Shadow
Copy Service (VSS) to define the backup application architecture in Windows,
which facilitates better association among multi-parties to work together on
backup services. The VSS provides administrators with the application program
interface (API) that is used to automatically create snapshots and perform
backups, allowing the third party's storage and backup software, and applications
to deal with storage backups with its assistance, which greatly reduces backup
time and complexity.


2013-7-29
Page 13 of 14

The VSS defines three different roles to describe which part each software plays
during performing backups. The three different roles are the writer, the requester,
and the provider. The writer describes the concrete object that uses data, such as
databases and mail servers. The requester focuses on the backup software, such
as the NetBackup of Veritas and the NTBBackup of Windows. The provider is the
actual provider of the snapshot. It can be an operating system, LUN management
software, and hardware, such as arrays. Take NBU backup for example. The
realization process is displayed in Figure 7.
Figure 7 The VSS invoking HyperClone to perform backups

The backup software NBU sends a backup request to the VSS, assigning
the writer that needs a backup.
After the database server that plays a writer receives the forwarded backup
request from the VSS, it will flush disks in the buffer and report to the VSS.
The VSS checks whether the array that plays the provider has the master
and slave LUN (the default priority of the provider is the hardware provider,
the software provider, and the system provider), and then the VSS will notify
the NBU to send the command of preparing for synchronizing the master
and slave LUN of the HyperClone.
In split mode, the NBU notifies the VSS to synchronize the master and slave
LUN of the HyperClone.
The VSS will temporarily lock the host services to avoid read or write
operations that will result in inconsistent data at the split time.
After the synchronization is complete, the VSS will forward the command of
dividing the slave LUN of the HyperClone to the OceanStor storage array.
The VSS notifies the host to start when the split of the HyperClone is
complete.
The NBU reads data from the slave LUN of the OceanStor storage array to
perform backups and then write backup data onto tapes.


2013-7-29
Page 14 of 14

In the host end, the VSS provider configures files to store information about the
master and slave LUN of the HyperClone, and assign the device IDs of the
master and slave LUNs, and snapshot types such as virtual snapshot or
HyperClone.
Apart from the NBU, backups can also be realized through BE. Both of them can
effectively avoid long backup time and support online data backups.

Huawei HyperClone Technical White Paper PDF

Transféré par

Informations du document

Description originale:

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Huawei HyperClone Technical White Paper PDF

Transféré par

Droits d'auteur :

Formats disponibles

Doc.

Vous aimerez peut-être aussi