R07_VNXUnified_SISTA_M07 - Advanced Storage Concepts

Copyright © 2011 EMC Corporation. All Rights Reserved.

Module 7: Advanced Storage Concepts

Upon completion of this module, you should be able to:

• Define metaLUNs

• Describe LUN migration

• Describe Fully Automated Storage Tiering (FAST)

• Describe Data Compression

• Describe FAST Cache

• Describe LUN Shrinking

Advanced Storage Concepts 1



During this lesson the following topics are covered:

• Creating a metaLUN

• Expanding metaLUNs

Lesson 1: Configuring metaLUNs


Copyright © 2011 EMC Corporation. All Rights Reserved. Advanced Storage Concepts - 3

Creating a metaLUN with the Expand Storage Wizard

Right click and

select Expand


Expand Storage Wizard – Summary


Expand Storage Wizard – Results


Verifying metaLUN Expansion with Unisphere

Right click and

select Show

Component LUNs


Host LUN Expansion using diskpart

Original 50GB LUN

After running diskpart

100GB LUN created

After expansion but

before diskpart is

run


metaLUN Usage and Recommendations

• Use a metaLUN when storage consolidation in needed

• LUN expansion is a requirement

• Recommendations Bind component LUNs to match I/O type expected

Use default stripe element size ( 128 block )

Activate read and write cache using defaults

Ensure write aside cache for component LUNs is the default (2048)

Use RAID 5 group of at least four disks (4+1) and larger

Use RAID 1/0 groups of at least four drives (2+2) and larger

Use RAID 6 groups of eight drives (6+2) is planning to expand disk groups in the future

Do not use the component offset to adjust for stripe alignment metaLUNS have their own value


metaLUNs Expansion Scenarios and Guidelines

• Concatenation – happens “instantly”

no performance impact because no data must be moved

• Build from scratch – happens “instantly”

Create a new base LUN, then expand it

No performance impact, Flare knows there is no data to move

• Expand a LUN containing data – time depends on size of the base LUN

ALL data is read and striped even if only one byte was written

• Expand a LUN while app is running – time depends on size and workload

All data must be read and striped

Workload and meta expansion must share capacity

• Re-stripe time is independent of element size multiplier

• Re-striping requires read and write of entire LUN Linear relation to size of existing LUN

• Expanding more than one LUN on same RAID groups at a time is not recommended – much slower

Advanced Storage Concepts - 9




• Benefits of LUN migration

• Migrating a LUN

Lesson 2: LUN Migration



LUN Migration Guidelines and Benefits

• LUN Migration moves data from a source LUN to a destination LUN within a single storage system

Any RAID type to any RAID type, SAS to NL-SAS, EFD to SAS

Any LUN type (Thin, Thick or Traditional)

Same size or larger

No disruption to the applications running on the host

• LUNs can be the target or source of LUN migration operations

• Can enhances performance or increase disk utilization for your changing business needs and applications

allows the user to change LUN type and characteristics


LUN Migration Restrictions

• Neither LUN may be a private LUN or Hot Spare

• Neither LUN may be binding, expanding, or migrating

• Either or both may be metaLUNs

• Destination LUN may not be in a storage group

• Destination LUN may not be part of SnapView or MirrorView operations

• Destination LUN may be larger than Source LUN


LUN Migration Operation

• Data is copied from Source LUN to Destination LUN

Source stays online and accepts I/O

• Destination assumes identity of Source when copy completes

LUN ID, WWN

storage group membership

• Source LUN is unbound after copy completes

• The migration process is non-disruptive

There may be a performance impact

• LUN Migration may be cancelled at any point

Storage system returns to its previous state


Starting a LUN Migration

Select the source LUN

and click on migrate


Viewing LUN Migration State




• FAST functionality

• Benefits of FAST

• Configuring a heterogeneous storage pool

• Configuring FAST

Lesson 3: Fully Automated Storage Tiering (FAST)



FAST

• Software that manages the placement of data blocks on tiers in storage pools

Requires an enabler

• Automatically moves more active data to faster drives and less active data to slower drives

• Less active data can be moved from Tier 1 storage without users noticing the difference

• FAST eliminates issues with having to migrate LUNs between pools because a pool can be created that allows Flash (SSD), SAS, and NL-SAS, drives to exist in a single pool.


Initial Tier Settings for Pool LUNs

• Does not require a FAST enabler

• Optimize for Performance

No tier setting specified

• Highest Tier Available

Sets the preferred tier for data placement to the highest tier

• Lowest Tier Available

Sets the preferred tier for data placement to the lowest tier



FAST Policies

• Auto-Tier

Moves data to a tier based on the LUN performance statistics

• Highest Tier Available

Moves data to the highest tier available

Flash/SSD, SAS, and NL-SAS

• Lowest Tier Available

Moves data to the lowest tier available

• No Data Movement

Moves no data between tiers, and retains the current tier placement.



FAST Options for Storage Pools

Drives

LUNs

Storage

Pool

Mixed-Use Pools

Homogenous and Heterogeneous LUNs

Lowest

Available Tier

Highest

Available Tier

FLASH SAS NL-

SAS

FLASH SAS NL-

SAS Optimized auto-Tiering


FAST Relocation Schedule and Pools

• Array-wide Relocation Schedule allows the user to define the days and duration for relocations

• Each pool has the option of using the array-wide scheduler:

For pools existing before FAST is installed, the default setting is OFF

For pools created after FAST is installed, the default setting is ON

• Users can over-ride the setting as needed.

• All pools have a manual “relocate now” option.


Storage Pool Properties – Tiering Options


Use the dropdown to display

additional options such as

Manual


Starting and Stopping Data Relocation


Viewing Tiering Summary

Configured from the customize

option on the Dashboard




• Data Compression functionality

• Benefits of Data Compression

Lesson 4: Data Compression



Data Compression

• Reduce amount of storage needed by compressing data

• Compression is a LUN attribute

• All processing happens in the background

• No interruption to Host I/O

Decompress on host read

Host writes are not compressed in-line

• LUN can become less compressed over time due to host writes

• System kicks off compression automatically when LUN is less compressed

• Intended for relatively inactive LUN, such as the backup copy


Data Compression - Operation

• Initial compression processes the entire LUN All allocated blocks for thin LUNs

• Subsequent compression (recompression) triggered when system-defined uncompressed data threshold is reached Processes only uncompressed data

Threshold: 10% of user capacity or 10 GB – whichever comes first Triggered for 50 GB thin LUN after 5 GB of new data written since last compression

Triggered for 200 GB thin LUN after 10 GB of new data written since last compression

• Compression Process Processes data in 64 KB chunks. Writes compressed data to disk if it saves at least 8KB.

Otherwise, write uncompressed data to disk

Algorithm provides benefits of zero detection and data reduction

User settings for compression rate of High/Medium/Low

• Compressed LUNs are Thin Consumed capacity of compressed thin LUN reflects savings due to both Thin and

Compression


Data Compression – RAID Group and Pool LUNs

• System initiates a migration on behalf of the user

The migration is private, not managed by the user

Counts against LUN migration feature limits

RAID Group LUN is in migrating state

System hides migration for thick LUN and shows compressing state

• Destination pool

RAID group LUN: provided by the user

Thick LUN: the same pool where Thick LUN resides

• A thin destination LUN is created for the user in the destination pool

Slices are reserved in the pool to ensure LUN can be copied to pool

• Source LUN is unbound when initial compression is completed


Data Compression - Restrictions

• Cannot turn on compression for

Private LUNs (Write Intent Log, Clones Private LUNs, Reserved LUNs, member LUNs of metaLUN)

Snapshot LUN

LUN that is being migrated

Mirrored LUNs if one of the mirrored arrays runs on a bundle that does not support mirroring for Thin LUN (pre-R29)

Expanding or shrinking pool-based LUN

• Cannot turn on or off compression for destroying or offline LUN

• Cannot migrate a compressed LUN which has a private migration in progress

• NDU

Cannot NDU if compressed LUN is initializing or decompressing

Cannot turn compression on or off while NDU is in progress


Turning Compression on/off for a RAID Group LUN


Turning Compression on/off for a Thick LUN

Check the box to turn on

compression LUN Properties


Data Decompression - Overview

• Decompression “Undoes” compression Initiated by disabling the compression LUN property

System processes all thin LUN blocks and writes uncompressed data back to the LUN

User compression rate settings of High/Medium/Low valid for decompression

• If compression LUN property is turned off during initial compression Decompressed LUN is the original LUN

• If compression LUN property is turned off after initial compression Decompressed LUN preserves the “fullness” of the original LUN

thin LUN Decompressed LUN is a thin LUN

RAID group LUN or Thick LUN Decompressed LUN is a fully-provisioned thin LUN (holes will be filled with zeros)

• Decompression will pause if pool consumption reaches a system-defined threshold (~91%)

• User can re-enable compression during decompression


LUN Compression Reporting




• FAST Cache functionality

• Benefits of using FAST Cache

• Configuring FAST Cache

Lesson 5: FAST Cache



VNX FAST Cache

• FAST Cache extends the storage system’s existing caching capacity for a better system wide performance

Available with VNX OE 05.31 or later

Introduced on CX4 with FLARE 30

Maps frequently accessed data to flash drives

• FAST Cache acts as an application accelerator

Provides flash drive level performance to the data which is accessed most frequently

• Reduces the load on other storage system resources

• FAST Cache is a non-disruptive online process

Uses existing memory allocation interface and does not use host cycles


VNX FAST Cache – Mode of Operation

• Read/Write Mode

Flash drives are configured as RAID 1 mirror pairs

Application gets the acknowledgement for an IO operation

For write operations, the FAST Cache page is marked as dirty and an acknowledgement is sent back to the host

Design insures all workloads are spread evenly across all the flash drives that have been used for creating FAST Cache


Creating VNX FAST Cache

FAST Cache disk selection

window


Configuring FAST Cache on a RAID Group LUN

FAST Cache will be enabled by

default on a RG LUN if the FAST

Cache enabler is installed

FAST Cache can be configured

from the Cache tab under LUN

Properties on an individual LUN


Configuring FAST Cache for Pool LUNs

FAST Cache is set at the Storage Pool

level by default for all pool LUNs if

FAST Cache enabler is installed

FAST Cache can be enabled on

Storage Pools after the pool has

been created




• LUN Shrink functionality

• Benefits of using LUN shrink

• Configuring LUN Shrink

Lesson 6: LUN Shrink



Feature - LUN Shrink

• Reclaims unused storage space

• Shrinking a Volume

Shrinking the host accessible Volume space

Freeing underlying storage by reducing the capacity of a FLARE LUN or MetaLUN

• Microsoft Windows Server 2008 can now shrink Volumes

ONLY Operating System that supports LUN Shrink

Basic and Dynamic volumes support

Single Dynamic or Spanned Dynamic volumes support


LUN Shrink Supported Configurations (LUNs and MetaLUNs) • VNX Block OE supports

Multiple shrink request on same LUN

Parallel shrink requests on LUNs of same RAID Group

• Shrink requests are returned successful to host if RG expansions/defrags are in progress at that time

• MetaLUNs with any number of LUNs and component stripe width supported

Configured User Capacity reflects the Shrunk Size of the DISKRAID.exe shrink command

Total User Capacity and Total Physical Capacity are updated



Module Summary

Key points covered in this module:

• Defined metaLUNs

• Described LUN migration

• Described Fully Automated Storage Tiering (FAST)

• Described Data Compression

• Described FAST Cache

• Described LUN Shrinking

Documents

R07_VNXUnified_SISTA_M07 - Advanced Storage Concepts