3-1 Storage Resource Tuning Technologies and Applications
3-1 Storage Resource Tuning Technologies and Applications
Applications
Foreword
1 Huawei Confidential
Objectives
On completion of this course, you will be able to understand the service features,
implementation principles, and application scenarios of the following features:
SmartThin
SmartTier
SmartQoS
SmartDedupe
SmartCompression
SmartMigration
2 Huawei Confidential
Contents
1. SmartThin
2. SmartTier
3. SmartQoS
4. SmartDedupe
5. SmartCompression
6. SmartMigration
3 Huawei Confidential
Overview
⚫ The traditional deployment of a storage system has the following problems:
Adverse impact or even interruption on services when expanding the storage space
Uneven storage space utilization
Low storage efficiency
⚫ SmartThin can allocate the storage space on demand to improve storage resource
utilization and fully meet service requirements.
4 Huawei Confidential
• If the actual amount of data is larger than expected, LUN space can be adjusted dynamically.
As public space, free space can be allocated to any LUN that needs space. In this way,
storage space utilization and effectiveness are improved. In addition, LUN space can be
adjusted online without affecting services during capacity expansion.
▫ Improvement 1: Real space is not allocated to LUNs when LUNs are created, but is
allocated on demand when LUNs are being used.
DB
Data
RAID
server
Disk
DB
Data
RAID
server
Space Disk
DB
Data
server RAID
Disk
DB
Data
server RAID
5 Huawei Confidential
• SmartThin does not allocate all space in advance, but presents users a virtual storage space
larger than the physical storage space. In this way, users see a larger storage space than the
actual storage space. SmartThin allocates the space based on users’ demands. If the storage
space is insufficient, users can add back-end storage units to expand the system capacity.
The whole expansion process is transparent to users without system shutdown.
• SmartThin creates thin LUNs based on a RAID 2.0+ virtual storage resource pool, that is,
thin LUNs coexist with thick LUNs in the same storage resource pool. A thin LUN is a logic
unit created in a storage pool, which can be mapped to and then accessed by a host. The
capacity of a thin LUN is not an actual physical space but a virtual value. Only when the thin
LUN starts to process an I/O request, physical space can be applied from the storage
resource pool based on the capacity-on-write policy.
• SmartThin allows the capacity detected by a host to be larger than the actual capacity of a
thin LUN. The capacity detected by a host is the capacity that a user creates for a thin LUN,
namely the volume capacity (virtual space) displayed on a host after a thin LUN is created
and mapped to the host. The actual capacity of a thin LUN refers to the physical space
actually occupied by the thin LUN.
• In addition, SmartThin allows users to cerate a thin LUN whose capacity is larger than the
maximum available space of a storage pool.
SmartThin Read Process
1. A thin LUN receives a read request 2. Queries the mapping table between the thin
from a host. LUN and the storage pool.
3. Confirms that the space is allocated by the storage pool and returns the data read from
the corresponding area in the storage pool to the host.
Data
2 1 D 3
1 2 D D D
3 D
1 1 3 5 2
0000 4
2 4 6 5 D
3
Thin LUN 6 Storage pool
Mapping table
1. The thin LUN receives a read request from the host.
2. Queries the mapping table between the thin LUN and the storage pool.
3. Confirms that the space is not allocated by the pool and returns all zeros to the host.
6 Huawei Confidential
▫ After receiving a read request from a host, a thin LUN queries the mapping table
between the thin LUN and the storage pool to check whether an actual storage space
has been allocated to the thin LUN by the storage pool.
▫ If the storage pool allocates an actual storage space to the thin LUN, SmartThin uses
direct-on-time to read data from the actual storage space and returns the data to the
host.
▫ If the storage pool does not allocate an actual storage space to the thin LUN, no data
is written, SmartThin returns all zeros to the host.
• Direct-on-time
▫ When capacity-on-write is used, the relationship between the actual storage area
and logical storage area of data is not calculated using a fixed formula but
determined by random mappings based on the capacity-on-write principle. Therefore,
when a thin LUN is read or written, the relationship between the actual storage area
and logical storage area must be redirected based on a mapping table. A mapping
table is used to record the mappings between an actual storage area and a logical
storage area. A mapping table is dynamically updated during writes and is queried
during reads. Therefore, direct-on-time is classified into read direct-on-time and
write direct-on-time.
SmartThin Write Process
2. Queries the mapping table between the thin
1. A thin LUN receives a write request
LUN and the storage pool.
from a host.
3. Confirms that the space is allocated by the pool and performs the write process on the corresponding area
in the storage pool. If the write request asks for releasing space, the space is released.
Data
Data
2 1 D 3
1 2 D D
3 3
1 1 3 5 2
4
2 4 6 5
1. A thin LUN receives a write request
from the host. Storage pool
Thin LUN 6
2. Queries the mapping table between the thin
LUN and the storage pool. Mapping table
3. If the space is not allocated by the pool, the storage system allocates the space first. And then performs write process on the corresponding
area in the storage pool. If the write request asks for releasing space, a message is returned to the host.
7 Huawei Confidential
▫ Upon receiving a write request from a host, a thin LUN queries the mapping table
between the thin LUN and the storage pool to check whether an actual storage space
has been allocated to the thin LUN by the storage pool.
▫ If the storage pool has allocated an actual storage space to the thin LUN, data is
written to the corresponding area in the storage pool (based on direct-on-time). If
the write request asks for releasing space, the system releases the space and returns
a response to the host indicating a successful data write.
▫ If the storage pool does not allocate an actual storage space to the thin LUN,
SmartThin uses capacity-on-write to allocate an actual storage space from the pool,
uses direct-on-time to build a relationship between the actual storage space and
logical storage space, and writes data to the actual storage space. If the write request
asks for releasing space, a write success acknowledgement is directly returned to the
host.
• Capacity-on-write
▫ Upon receiving a write request from a host, a thin LUN uses direct-on-time to check
whether a physical storage area is allocated to a logical storage area provided for the
request. If a physical storage area is not allocated, a space allocation task is triggered,
and the grain size (minimum granularity) is 64 KB. Then data is written to the newly
allocated physical storage area.
Application Scenarios
⚫ SmartThin can help core system services that require high service continuity, such as bank
transaction systems, expand system capacity online without interrupting ongoing services.
⚫ SmartThin can assist with on-demand physical space allocation for services where the growth of
application system data is hard to be accurately evaluated, such as email services and web disk
services, preventing a space waste.
⚫ SmartThin can assist with physical space contention for mixed services that have diverse storage
requirements, such as carriers' services, to achieve optimized space configuration.
8 Huawei Confidential
Configuration Process
9 Huawei Confidential
Contents
1. SmartThin
2. SmartTier
3. SmartQoS
4. SmartDedupe
5. SmartCompression
6. SmartMigration
10 Huawei Confidential
Overview
⚫ The random distribution of hot and cold data fails to fully utilize disk characteristics of different media.
SmartTier automatically matches different active data with storage media of different characteristics. For
example, cold data is stored on NL-SAS disks, and hot data is stored on SSDs. In this way, data flows vertically,
improving storage system performance and reducing costs.
High-performance tier
SSDs
Performance tier
SAS disks
Most active data
11 Huawei Confidential
• With the development of disk technologies, storage systems support more types and a
growing number of storage media. Each type of storage media offers its unique advantages
and disadvantages in performance and cost, and it is difficult for users to strike a balance
between storage costs and storage performance.
• Data features
▫ Cold data: stored in or migrated to a capacity tier without any performance reduction
after migration.
Dividing Storage Tiers
⚫ In the same storage pool, a storage tier is a collection of storage media with the same performance. Each
storage tier respectively uses the same type of disks and RAID policy.
12 Huawei Confidential
• A storage pool is a logical combination of one or more storage tiers, supporting a maximum
of three storage tiers. The types of disks in a storage pool determine the allowed number of
storage tiers. A storage pool housing one type of disks can only create a single storage tier
and therefore does not support SmartTier for intelligent data storage management.
Three Phases for Implementing SmartTier
⚫ The storage system undergoes three phases of I/O monitoring, data placement analysis,
and data relocation to implement SmartTier.
Data placement analysis The data placement analysis module ranks the activity levels
of all data blocks.
Data migration The data migration module migrates data based on the
ranking result and data migration policies.
13 Huawei Confidential
• If a storage pool contains more than one type of disks, SmartTier can be used to fully utilize
the storage resources. During data migration, a storage pool identifies data activity levels
by data blocks and migrates complete data blocks to another storage tier.
• I/O monitoring and data placement are automatically performed by the storage system,
and data migration is initiated manually or by a user-defined or scheduling policy.
SmartTier Key Technologies
Migration Monitoring
Initial capacity Data
policy statistics
allocation migration
formulation analysis
14 Huawei Confidential
Application Scenarios
15 Huawei Confidential
• Since cold data is stored on NL-SAS disks, the storage space of the high-performance SSDs is
freed up for hot data. SSDs provide hot data with quick response and high IOPS. In this way,
the overall storage system performance is aggressively improved.
Configuration Process
Storage system-level configuration includes the configuration
Start
of data migration speed, which is applied to all storage pools in
a storage system.
Check the license. Storage pool-level configurations include configurations of
data migration granularity, RAID policy, data migration plan,
enabling I/O monitoring, and forecast analysis. The mentioned
Configure SmartTier parameters based on
the storage system level. configurations are applied to a single storage pool.
End
16 Huawei Confidential
• The configuration process of SmartTier in the storage system includes checking the license,
configuring SmartTier parameters based on the storage system level, configuring SmartTier
parameters base on the storage pool level, and configuring SmartTier parameters based on
the LUN level.
Contents
1. SmartThin
2. SmartTier
3. SmartQoS
4. SmartDedupe
5. SmartCompression
6. SmartMigration
17 Huawei Confidential
Overview
18 Huawei Confidential
• When multiple applications are deployed on the same storage device, users can obtain
maximized benefits through the proper configuration of SmartQoS.
FIFO 7
8 4
7 1
9 8
6
5 6
4 2
3 9
2 5
1 3
19 Huawei Confidential
• The I/O priority scheduling technology of SmartQoS is implemented based on LUN priorities.
• Each LUN or file system has a priority property, which is configured by a user and saved in a
database. When a host sends an I/O request to a storage array, the storage array gives a
priority to the I/O request based on the priority of the LUN or the file system that will
process the I/O request. Then the I/O carries the priority throughout this processing
procedure.
• When a LUN or a file system is created, its I/O priority needs to be specified. If not, the LUN
or the file system is granted the low priority by default.
• After a LUN or a file system is created, its I/O priority can be manually changed.
I/O Traffic Control
Application server 1 Application server 2
I/O I/O
1 request 1 request
20 Huawei Confidential
• The performance goal of I/O traffic control is achieved based on token distribution and
control. When a user sets a performance upper limit for a traffic control group, the upper
limit is converted into the number of corresponding tokens. In a storage system, If the IOPS
is limited, an I/O corresponds to a token. If the bandwidth is limited, a token is allocated to
a sector.
• Each traffic control queue has a token bucket. SmartQoS puts a certain number of tokens
into the token bucket of each traffic control queue periodically. The number of tokens is
determined by the performance upper limit set for the traffic control group. For example,
if the performance upper limit is set to IOPS = 10,000, the token distribution algorithm sets
the maximum number of tokens in the token bucket to 10,000 for the traffic control group.
• When processing a traffic control queue, the queue checks whether the token bucket has
robust tokens. If yes, one I/O is processed, and its corresponding token is consumed. If no,
SmartQoS does not process I/Os in this queue until there are tokens in the token bucket.
Application Scenario
User Type Service Quality
Requirements
Subscriber A (gold subscriber) High
21 Huawei Confidential
• SmartQoS allows users to create different SmartQoS policies. In this way, the service
running and quality of high-level users are preferentially ensured when resources are
insufficient.
• For cost reduction, some users will not build their dedicated storage systems independently.
They prefer to run their storage applications on the storage platforms offered by storage
resource providers. This lowers the total cost of ownership (TCO) and ensures the
application continuity. On such shared storage platforms, applications of different types
and features content for storage resources, so the high-priority users may fail to obtain
their desired storage resources.
Configuration Process
Start
22 Huawei Confidential
Contents
1. SmartThin
2. SmartTier
3. SmartQoS
4. SmartDedupe
5. SmartCompression
6. SmartMigration
23 Huawei Confidential
Overview
⚫ SmartDedupe eliminates redundant data from a storage system and reduces the physical
storage capacity for storing data to meet the increasing needs for storage capacity.
⚫ Dorado V6 storage systems support online deduplication and post-process similarity
deduplication.
Online deduplication: Data is deduplicated before being written to disks.
Post-processing similarity deduplication: Data is written to disks in advance and then read and
deduplicated when the system is idle.
24 Huawei Confidential
Working Principle of Online Deduplication
Data to be deduplicated
25 Huawei Confidential
• 2. The storage system uses the weak hash algorithm to calculate the fingerprint information
about a data block that is newly written into the storage system.
• 3. The storage system checks whether the fingerprint information about the newly data
block is consistent with the fingerprint information in the fingerprint library.
▪ Same=>Old block
▪ Different=>New block
▫ If no, the newly written data is a new block.
• 4. For old block: The storage system maps its fingerprint information and storage location
mapping to the existing data block in the fingerprint library.
• 5. For new block: The storage system writes new data into the disk and records its
fingerprint information in the fingerprint library and mapping in the storage location.
Working Principle of Post-processing Similarity Deduplication
Data to be deduplicated
Differentially
Writes data. +1
compresses.
26 Huawei Confidential
• 1. A storage system divides newly written data into blocks. The default data block size is 8
KB.
• 2. The storage system uses the similar fingerprint algorithm to calculate the similar
fingerprint information about a newly written data block.
• 3. The storage system writes a data block to the disk, and writes its fingerprint and location
information to the opportunity table.
• 4. The storage system periodically checks whether there is similar fingerprint information in
the opportunity table.
▫ If yes, goes to 2.
• 5. The storage system checks whether similar blocks are the same through the result of the
byte-by-byte comparison.
▫ If they are same, the storage system deletes this data block, and maps its fingerprint
information and storage location to the existing data block.
▫ If they are similar, the storage system performs differential compression on data
blocks, records its fingerprint information to the fingerprint library, updates the
fingerprint information to the metadata of data blocks and recycles the spaces of
these data blocks.
Application Scenarios of SmartDedupe
⚫ Virtual Desktop Infrastructure (VDI) is a common application scenario of deduplication.
⚫ In VDI applications, users create multiple virtual images on a storage device. These images
have a large amount of duplicate data. As the amount of duplicate data increases, the
storage system space fails to meet service running requirements. SmartDedupe can delete
duplicate data between images to release storage resources and store more service data.
27 Huawei Confidential
Configuration Process
Start
End
28 Huawei Confidential
Contents
1. SmartThin
2. SmartTier
3. SmartQoS
4. SmartDedupe
5. SmartCompression
6. SmartMigration
29 Huawei Confidential
Overview
⚫ SmartCompression reorganizes data to save storage space and improves the data transfer,
processing, and storage efficiency under the precondition that no data is lost. The storage
system supports online compression, that is, only newly written data is compressed.
⚫ The storage systems of the Dorado V6 storage systems support online compression and
post-compression, both of which are lossless compression.
Online compression: Data is compressed before being written to disks.
Post-compression: Data is written to disks in advance and then read and compressed when the
system is idle.
30 Huawei Confidential
Working Principle of SmartCompression
Compression window
Data to be
... ... abcdefg abc hj abchj
compressed
(LZ77 example)
31 Huawei Confidential
Application Scenarios of SmartCompression
Engineering
File and
Databases
Services Seismic Geological
Data
32 Huawei Confidential
• Database: A database is the optimal application scenario for data compression. Many users
would like to save more than 65% storage space at the expense of slight performance
deterioration.
• File service: The file service is a common application scenario for data compression. For the
storage system with the file service enabled, peak hours occupy half of the total service
time and the dataset compression ratio of the system is 50%, in this case,
SmartCompression slightly decreases the IOPS.
• Engineering data and seismic geological data: Features of engineering and seismic
geological data are similar with the those of database backup data. This type of data is
stored in the same storage format, but there is few duplicate data. Therefore, such data can
be compressed to save the storage space.
Scenarios Where SmartDedupe and SmartCompression Are Used
Together
⚫ The deduplication and compression technologies can be used at the same time to achieve
the optimal space saving effect.
⚫ Application scenarios:
VDI and VSI scenarios
Less storage Longer
Less cost
Data tests or development systems space SSD life
33 Huawei Confidential
• SmartDedupe can combine with SmartCompression for data tests or development systems,
storage systems with the file service enabled and engineering data systems.
• VDI is usually applied for creating multiple virtual images which have a large amount of
duplicate data at a singe storage device. As the duplicate data increases, a storage system
may fail to work properly. SmartDedupe and SmartCompression can prevent such
circumstance.
• Advantages:
▫ Procurement cost: In the initial phase, a smaller number of storage devices can be
purchased to store the same amount of data.
▫ TCO: Less management manpower is needed because fewer storage devices are
purchased. In addition, costs in room space, power consumption, cooling, and O&M
drop accordingly.
Start
2. Enabling
SmartCompression for Enable SmartCompression.
LUNs
End
34 Huawei Confidential
Contents
1. SmartThin
2. SmartTier
3. SmartQoS
4. SmartDedupe
5. SmartCompression
6. SmartMigration
35 Huawei Confidential
Overview
⚫ SmartMigration is a key service migration technology. Services on a source LUN can be
completely migrated to a target LUN without interrupting host services. The target LUN
can totally replace the source LUN to carry services after the replication is complete.
36 Huawei Confidential
• "Completely" means that after the service migration is complete, all service data has been
replicated from a source LUN to a target LUN.
• SmartMigration features:
▫ Reliable service continuity: supports migrating service data online to prevent any loss
caused by service interruption during service migration.
▫ Stable data consistency: timely synchronizes data changes on a host to both source
LUN and target LUN during service data migration, ensuring data consistency after
migration and preventing data loss.
37 Huawei Confidential
• A storage system uses the virtualized storage technology. Virtual data in a storage pool
consists of metadata volumes and data volumes.
▫ Metadata volumes: record the data storage locations, including IDs of LUNs and data
volume IDs. IDs of LUNs are used to identify LUNs and data volume IDs are used to
identify physical space of data volumes.
38 Huawei Confidential
• Pair: In SmartMigration, a pair indicates the data migration relationship between a source
LUN and a target LUN. A pair can have only one source LUN and one target LUN.
• The two synchronization modes of service data are independent and can be performed at
the same time to ensure that service data changes on the host can be synchronized to the
source LUN and the target LUN.
▫ The LM module writes the data to the source LUN and target LUN and records write
operations to the log.
▫ The source LUN and target LUN return the data write result to the LM module.
▫ The LM module determines to clear LOG or not based on the write I/O result.
39 Huawei Confidential
• LUN information exchange is the prerequisite for a target LUN to take over services from a
source LUN after service information synchronization.
• In a storage system, each LUN and its corresponding data volume have a unique identifier,
namely, the ID of a LUN and data volume ID. A source LUN corresponds to a data volume.
The former is a logical concept whereas the latter is a physical concept.
• Before LUN information exchange: A host identifies a source LUN by the ID of the source
LUN. The ID of a LUN corresponds to a data volume ID.
• During LUN information exchange: A source data volume and a target data volume ID are
exchanged. The physical storage space to which the source LUN points becomes the target
data volume.
• After LUN information exchange: The ID of the source LUN is unchanged, and users sense
no fault because services are not affected. The ID of the source LUN and target data volume
ID form a new mapping relationship. The host actually read and writes physical space of the
target LUN.
SmartMigration Pair Splitting
⚫ Splitting is performed on a single pair. The splitting process includes stopping service data
synchronization between the source LUN and target LUN in a pair to exchange LUN information,
and removing the data migration relationship after the exchange.
1. Information
exchange
2. Pair splitting
40 Huawei Confidential
• In splitting, host services are suspended. After information is exchanged, services are
delivered to the target LUN. In this way, service migration is transparent to users.
• Pair splitting: Data migration relationship between a source LUN and a target LUN is
removed after LUN information is exchanged.
▫ After the pair is split, if the host delivers an I/O request to the storage system, data is
only written to the source LUN.
▫ The target LUN stores all data of the source LUN at the pair splitting point in time.
▫ After the pair is split, no connections can be established between the source LUN and
target LUN.
• The consistency splitting of SmartMigration means that multiple pairs exchange LUN
information at the same time and concurrently remove pair relationships after the
information exchange is complete, ensuring that data consistency at any point in time
before and after the pairs are split.
• In scenarios where multiple pairs are used, such as in medium- and large-size database
applications, data, logs, records, and other files are stored on LUNs that are associated with
one another in a storage system. Splitting cannot ensure that information in one LUN is
always associated with that in another. If data in a LUN is unavailable, data in the other
LUNs may become invalid. Consistency splitting is used to ensure data consistency.
Configuration Process
Start
End
41 Huawei Confidential
SmartTier
SmartQoS
Definitions, working
Storage resource tuning technology
principles, and configuration
and application processes
SmartDedupe
SmartCompression
SmartMigration
42 Huawei Confidential
Quiz
1. (True or false) SmartTier cannot be enabled for a storage pool whose member disks are of the same type. ( )
2. (Multiple-choice) Which of the following migration policies can be set for LUNs? ( )
A. Automatic migration
D. No migration
43 Huawei Confidential
• Answers:
▫ 1. T
▫ 2. ABCD
Quiz
3. (Single-answer question) Which status must a pair be before consistency splitting during LUN
migration? ( )
A. Migrating
B. Stop
C. Normal
D. Migrated
44 Huawei Confidential
• Answers:
▫ 3. D
More Information
Enterprise Huawei
technical support enterprise
app business app
45 Huawei Confidential
▫ Covers all popular product documents, cases, and bulletins of Huawei. Users can
quickly query commands, alarms, and spare parts, and scan the QR to view the
device information and simple as well as intuitive video guide. This app provides
uninterrupted enterprise technical support.
▫ Provides one-stop mobile ICT portals for customers and partners to understand
Huawei's comprehensive product and solution information in the enterprise ICT field
anytime and anywhere.
Recommendations
⚫ Popular tools
HedEx Lite
Network Documentation Tool Center
Information Query Assistant
46 Huawei Confidential
• Popular tools:
▫ HedEx Lite: Huawei product document management tool, which allows users to
browse, search for, update, and manage product documentation.
▫ Network Documentation Tool Center: The documentation tool for network products
is a good assistant for bidding support, network planning, project delivery, and
upgrade and maintenance.