Storage for Networking Professionals

8/2/2019 Storage for Networking Professionals

1/75

Page 1

Storage for Networking Professionals

Presented by

Elaine Silber

Training and Certification Director

Infinity I/Owww.InfinityIO.com


2/75

Page 2Storage for Networking Professionals Copyright 2004 Storage Networking Industry Association. All Rights Reserved.

SNIA Legal Notice

The material contained in this tutorial is copyrighted by theSNIA.

Member companies and individuals may use this material inpresentations and literature under the following conditions:

Any slide or slides used must be reproduced without modification

The SNIA must be acknowledged as source of any material usedin the body of any document containing material from thesepresentations.

This presentation is a project of the SNIA EducationCommittee.


3/75


Aims and Objectives

This tutorial is intended to be an introduction to

basic storage technologies, techniques andterminology associated with data storage devicesfor networking professionals.

It provides an introduction for industryprofessionals who wish to acquire thesefundamentals of data storage before attendingmore targeted networked storage presentations


4/75


Outcomes

After attending this tutorial you should be able to: Understand basic storage terms and technology and the basic

operation of hard disk drives Identify the components of the SNIA Shared Storage Model

Appreciate how disk drive characteristics impact performance

Describe storage concepts, including LUN mapping, zoning,volume manager and file systems

Describe basic storage protection techniques - RAID

Identify tape storage technologies

Learn Storage connectivity approaches

Understand Tiers of Storage (TOS) as seen from an InformationLife Cycle Management point of view


5/75


Networking vs Storage Language

The Language of Networking The Language of Storage

Data over Distance Data over Time

Information movement Information repository

Configuring and segmentingnetwork topologies

Formatting and partitioning harddisks, Tiers of Storage (TOS)

Hubs, bridges, switches, routers,gateways, NIC (Network InterfaceCard), HBA (Host Bus Adaptor)

JBOD, RAID, Tape driveControllers

Inherently error prone and

acceptable

Designed to maximize correct

delivery. Expect error free.

Client/Server applications Initiator/Target Functions

Email, Web Browsing, File

Management Network Interfaces

Backup, Archive, Mirror, Block

Management I/O interfaces


6/75


Networking vs Storage Language

The Language of Networking The Language of Storage

Speed, Solid State Capacity, Mechanical Movement

Media - Copper, Fiber, Wireless Media Disk, Tape

Bits in packets, CRC error checksevery packet

Bytes in blocks, Parity check everybyte

Buffers and Link transfer rates Kb/s, Mb/s, Gb/s

Cache and Disk/Tape I/O interfacetransfer rates MB/s Throughput

Interconnectivity network and

device latencies bottlenecks

Disk/Tape seek times and

latencies ms, us, nsApplications HTTP, NFS, SMTP

Transport Protocols

(TCP/IP/Ethernet), Fibre Channel

Logical Device Protocols SerialSCSI, FC, IDE/ATA

Physical Protocols Parallel SCSI


7/75


Historical View of Storage/Network

Host-attached storage

Network Administrators saw storage as a black box

Storage Managers saw the network as a cloud

NetworkAdmin

Storage Admin

Database Manager


8/75


Current View of Shared Storage

Storage is a shared resource

Separate from the computer system (host)

Hosts are consumers of storage


9/75


Overall Terminology Confusion

Classic Storage Model

Applications

Host

Disk

array?

Datamover?

Appliance?

JBOD?

Storage Domain:Anything goes?

Network?


10/75


SNIA Shared Storage Model

File/rec

ord

layer

Block

layer

Storage

Devices

Device

Host

StorageDo

main

Files/DatabasesFiles/Databases

Packing manysmaller things into afew larger ones..

Block LayerBlock Layer

Storage Devices disk drives, tapedrives, solid statedisk

Block

A

ggregation

Network

Applications

Database(dbms)

File system(FS)

Block Aggregation address mapping,concatenation,striping, mirroring


11/75

Page 11

Storage Basics

Disk Physical

Characteristics


12/75


Disk Drives Walk the Same Walk

From a component perspective, all disk drives

are basically the same. A disk platter, usuallycomprising an aluminum or glass substratematerial that is sputter-coated with a

magnetically corrosive media (chromium,ruthenium, etc.), provides a precision surfacearea in which discrete data bits can be written

into a series of tracks.

Jon William Toigo


13/75


The ABCs of Physical Addressing

Head

Read/write head

Defines a single disk surface

Cylinder

The information that can be accessedon a disk drive by all the heads, withouthaving to seek

Sector A subdivision of a disk surface which is

created during formatting (typically a

512-byte segment)


14/75


16 Logical Heads

The ABCs of Physical Addressing

Specifications on label:

1416 Cylinders (1416 tracks)

16 Heads

63 SectorsBut 4 Physical Heads


15/75


The ABCs of Disk Performance

Data transfer time is determined by:

Speed of I/O technology (SCSI, Fibre Channel, etc) Seek time (time for heads to move to a new track)

Latency (time for sector to rotate under heads)

Speed and size of disk drive buffer memory

Whether the demand equals offered load (Can drive keep up with it speed of disk drive vs throughput)


16/75

Page 16

Storage Basics

Storage Interfaces


17/75


Drives Do Not Talk the Same Talk

Drive Interface Parallel IDE/ATA or Serial ATA (sometimes

abbreviated SATA) Integrated Drive Electronics (IDE), AT Attachment (ATA)

Parallel SCSI or SAS (Serial Attached SCSI)

Fibre Channel


18/75



Drive Interface Parallel IDE/ATA

lower-cost devices than SCSI,

controller/drive electronics wedded to the drive itself.

IDE standardizes how disk drives are connected to servers

via a ribbon cable and an interface connector on a PC orserver motherboard.

IDE/ATA protocol five+ revisions,

For 10 years now speeds and feeds keeping pace with

speed- and capacity-hungry applications. Little improvement in transfer rate expected in future

Difficult to create arrays with Parallel ATA


19/75



Drive InterfaceSerial ATA (sometimes abbreviated SATA)

is a standards-based interface that improves on parallelIDE/ATA

interface transfer rate is 150Mb/s, (compared with parallelATA's 100Mb/s)

That's a 50% increase, particularly useful in applicationswith large data volume requirements, such as videoediting.

First Generation SATA drives not yet more efficient than

Parallel ATA, SATA-2 and SATA-3 - expected transfer rates of 300 and 600

Mb/s Still limited by max speed of PCI bus at 133 Mb/s


20/75



Drive Interface Parallel SCSI

Several standards-based iterations, Confusing descriptors Fast, Wide, Ultra Currently, UltraSCSI 3 or Ultra160 SCSI

transfer rates of 160Mb/s

attach up to 16 devices on a single bus. Ultra320 andUltra640 are demonstrated at trade shows

Doubling and quadrupling of transfer rates

No significant improvement in device attachment orcabling distance characteristics. Note Tape Drive speeds do not come close to SCSI bus

speeds which is a good thing in order to keep drives moving


21/75



Drive Interface SAS (Serial Attached SCSI)

Greater throughput

Higher device attachment capabilities

Greater cabling distances

http://www.scsita.org/aboutscsi/presentations.html


22/75



Reliability

SCSI Drives

Mechanical pieces of a SCSI diskdrive are manufactured with ahigher duty cycle Enterprisedrives, supporting high-

performance transactionprocessing databases, areconstantly being beaten on byrandom seeks.

IDE/ATA drives Considerably lower duty cycles.

For PCs.

CPU Utilization

SCSI & FC drives

Responsibility is split between thehost processor and aspecialized processor on thehard disk assembly for

performing functions such asqueue management.

IDE/ATA drives The electronics in an IDE/ATA

drive rely on the hostprocessor to perform allstorage tasks.


23/75

Page 23

Storage Basics

Protocols


24/75


SCSI Standards Architecture*

SCSI BlockCommands

(e.g., disk drive)

(SBC, SBC-2)

Reduced BlockCommands

(e.g., disk drive)

(RBC, RBCAM-1)

SCSI StreamCommands

(e.g., tape drive)

(SSC, SSC-2, SSC-3)

SCSI MediaChanger

(e.g., jukebox)

(SMC, SMC-2)

Multi-MediaCommands(e.g., DVD)

(MMC, MMC-2-5)

SCSI EnclosureServices

(SES, SESAM1,

SES-2)

Object BasedStorage Device

(OSD)

ManagementServer Commands

(MSC)

Primary Commands for All Devices(SPC, SPC-2, SPC-3)

SCSI ControllerCommands(e.g., RAID)

(SCC-2)

Device

SpecificCommandSets

Shared Command SetArchitectural Model

SAM

SCSIArchitecturalModel

(OriginallySCSI-3)

SCSI ParallelInterface

(SPI-2-5)

RelatedStandards and

Technical Reports(SDV, PIP, SSM,SSM2 and EPI)

Serial BusProtocol

(SBP-2-3)

IEEE-1394

FibreChannel

Protocol(FCP, FCP-2,

FCP-3)

FibreChannel

(FC)

SSA-SCSI-3Protocol

(SSA-S3P)

SSA-TL2

SSA- PH1 orSSA-PH2

SCSI-RDMAProtocol

(SRP, SRP-2)

InfiniBand TM

iSCSI

Internet

SerialAttached

SCSI(SAS, SAS-

1.1)

TransportProtocols

PhysicalInterfaces

*See TutorialSCSI The Protocol for all Storage Architectures


25/75


ABCs of SCSI Addressing

Data/Address Bus

Control Signals

Interface Interface InterfaceSCSIRAID

ID6

SCSIRAIDID0

SCSIRAID

ID4

SCSIInitiator

Host BusAdaptor

ID7

LUN 0LUN 1

LUN 2LUN 3

LUN 0LUN 1

LUN 2LUN 3

LUN 0LUN 1

Address = BUS : Target ID : LUN


26/75


How do you route SCSI?

SCSI isnt a routable protocol so how do you route SCSI?

SCSI-FCPSOF

Payload: SCSI Command in IUPayload: SCSI Command in IUCRC

FCFrameHeader

EOF

FCIP EthernetheaderIP

headerTCP FCIP

Header

SOF


EOF

FCS

FCFrameHeader

Ethernetheader

IPheader

TCP iFCPHeader

SO

F

FCFrame

Header


EO

F

FCS

iFCP

Ethernetheader

IPheader

TCPiSCSI

header Payload: SCSI Command PDUPayload: SCSI Command PDUFCSiSCSI


27/75

Page 27

Storage Basics

Basic Storage Protection


28/75


RAID Flavors

Why RAID? -Redundant Array of Independent Disk Original work was to get away from large monolithic disks

Better performance with more individual disks and smaller

RAID 1Mirrored Volumes

RAID 0+1Mirrored Array

RAID 4Block-Level Striping with Parity Disk

RAID 5Striping with Distributed Parity

RAID 10Mirrored Striping Array

See Appendix for further definitions


29/75


RAID Flavors


30/75


Software RAID

Volume management performed by server

Parity computation performed by server increased overhead RAID performance dependent on server performance and CPU load

For simple environments with lower performance and availabilityrequirements

Volume n

Volume,me B

Volume A

SERVER RAID STORAGE

Block A3 Parity 3

Block A2 Block B2

Block A1 Block B1

Block C3

Parity 2

Block C1

Block D3

Block D2

Parity 1


31/75


Hardware RAID

Volume management performed by RAID controller card

Embedded processor in RAID controller to reduce server overhead Parity computation performed by auxiliary processor in controller

Dedicated cache memory increases server write performance

Volume n

Volume,me B

Volume A

SERVER RAID STORAGE

Block A3 Parity 3

Block A2 Block B2

Block A1 Block B1

Block C3

Parity 2

Block C1

Block D3

Block D2

Parity 1


32/75


Future for RAID

A SNIA Technical WorkingGroup (TWG) is currently a

common RAID Disk DataFormat (DDF) Specification

Would allow storing RAIDconfiguration on physical

disks in a DDF by differentvendor implementations ina common format

Why? -- Would enable

data-in-place migrationamong systems fromdifferent vendors.

www.snia.org/tech_activities/ddftwg


33/75

Page 33

Storage Basics

Logical View

of Storage

Physical vs. Logical


34/75


Data Presentation Chain

RAID Controller

Phys. Blocks (C/H/S)

FileFile

presents

presents

Application

VolumeFileFile presents

Application

Users Applications

Applications Files SystemRecordstuples tables

File System VolumesMetadata

tables tablespaces

I/O SubsystemVolume Logical Blockspresents

Logical Blocks presents

Disk Drives

RAID Controller presents


35/75


What the Host Sees

SAN

?JBOD (Just a Bunch of Disks)

Identifying storage volumes (SCSI IDs and LUNs)

Block Access

Wh h H S i h JBOD


36/75


What the Host Sees with JBOD

Logical Disks Physical Disks

5 x 1GB

SAN 1GB1GB

1GB

1GB

1GB

SCSI LUNs(LogicalUnits)

5 x 1GBHost A

JBODDrive D (2GB)Drive E (500MB)

Drive F (500MB)

SCSI ID 3

SCSI ID 4

SCSI ID 1

SCSI ID 2

SCSI ID 0

LUN 0

LUN 0

LUN 0

LUN 0

LUN 0

Host B

Drive D (1GB)

Drive E (1GB)

Volume Manager HBA (Host Bus Adaptor)Utility

Wh h H S i h RAID


37/75


What the Host Sees with RAID

Logical Disks

Volume Manager HBA Utility RAID Configuration Utility

SAN3GB

1GB

1GB

Host A

Drive D (3GB)

Host B

Drive D (1GB)

Drive E (1GB)

SCSI LUNs2 x 1GB1 x 3GB

Physical Disks

5 x 1GB

RAID (Redundant Array

of Independent Disks)

SCSI ID 2

SCSI ID 0

SCSI ID 1

LUN 0

LUN 2

LUN 1

LUN 0

LUN 0

St Vi t li ti t W k


38/75


Storage Virtualization at Work

Logical Disks Physical Disks

10 x 1GB

SAN

3GB

1GB

1GB

Host A

Drive D (5GB)

5GB

4GB

1GB

Virtualized LUNs

1 x 1GB1 x 4GB1 x 5GB

Virtualization Software

MetadataManager

SCSI LUNs4 x 1GB2 x 3GB

RAID

Host B

Drive D (1GB)

Drive E (4GB)RAID

3GB

1GB

1GB

HBA UtilityVolume Manager RAID Configuration Utility

SCSI ID 2

SCSI ID 0

SCSI ID 1

LUN 0

LUN 0

LUN 0

LUN1

LUN 2

LUN0

LUN0

LUN 0

LUN 1

LUN 2

S WYSIWYG? N t


39/75


Summary: WYSIWYG? Not..

What the file systemmakes the userbelieve.

Myfile.doc

Myfile.xls

Myfile.ppt

1

2

1 0 4

5 9

6

8 3 7

1 3

1 2 1 4

1 5

1 1

Volume Manager keeps trackwhere data blocks are writtenReality


40/75

Page 40

Storage Basics

Tape (no less important)

Tape in the


41/75


p

SNIA Shared Storage Model

File/re

cord

layer

Block

layer

Storage

Devices

Device

Host

Block

Aggregation

Network

Applications

Database(dbms)

File system(FS)

Tape Appl icat ion

(e.g. backup soft w are)

Tapefo

rmat

system

Device

HostNetwork

Extentaggregation

Tape Devices

Tape Media

Why Tape?


42/75


Why Tape?

Speed Faster than disk? Scalability

Real estateGB/sq. ft.

Proven longevity Legal, Archiving Unlike disk, tape media and devices are not co-dependent

Transportableoff-site archiving and storagedisaster recovery Media could be read on compatible tape mechanisms

Mechanism failure does not compromise data Serial access is appropriate for reading and writing long streams

of data (but not for real-time backup) Disk to disk copy in theory, but disk to disk to tape is reality for

backup.

Question If an appliance controls the backup to tapes, can the backup

software read it?

ABCs of Tape Media


43/75


ABCs of Tape Media

Tape ReelTape Reel

Pinch RollersPinch Rollers

Tape ArmsTape Arms

Read/write headRead/write head

Take-up reelTake-up reel

Read/write headRead/write head

Leader pinLeader pin

Take-up reelTake-up reelTape ReelTape Reel

Media Compatibility


44/75


Media Compatibility

Tape drives and cartridges are typically developedtogether as inseparable technologies:

New tape drives may have the capability to readexisting cartridges made on older drives

Older drives may not have the capability to read the

cartridges written by newer drives

Tape Recording Method


45/75


Tape Recording Method

Data is written to tapes in large contiguous sequentialblocks (variable size)

Access time to a random block on a tape may be longdue to serial recording method:

Some technologies read tape from the middle outreduces

access time by 50% on average Tape can achieve streaming speeds in the range of

5 to 30MB/sec (uncompressed)

Compressed speeds vs uncompressed speeds Compression typically done at tape drive

Compression ratio is dependent on the data type(e.g. BMP vs JPEG format)

Automatic Tape Subsystems


46/75


Automatic Tape Subsystems

Tape Library Tape automation product

Requires host level specialized management software Management path different from data path

Automates media selection and loading

Uses bar code readers to facilitate media management

No tight connection betweentape library and tape drives(unlike tape drives and

tape cartridges)


47/75

Page 47

Storage Basics

Storage Connectivity

Storage Connectivity (DAS)


48/75


Storage Connectivity (DAS)

Traditional way of implementing storage

Storage is managed by a single host

Other hosts must access the storage through that singlehost, over the LAN

Direct-attached

internal storage

Direct-attached

external storage

Tape device

Direct-attached

external storage

SCSI

Fibre Channel

Storage Connectivity (NAS)


49/75


Storage Connectivity (NAS)

Provides access to storage over the Network

NAS devices contain a thin server that provides fileservices to other hosts on the LAN using network fileaccess methods

File access

NAS appliance

Storage Connectivity (SAN)


50/75


Storage Connectivity (SAN)

Servers and stand-alone storage devices, connected bya dedicated network

Any server can be configured to access any storagearray and/or storage to storage access

Servers and storage can scale independently Block level access

Storage

array

Tapedevice

Server

Storage ConnectivityT t T h l i


51/75


Transport TechnologiesParallel SCSI

FibreChannel(Serial)

iSCSI(Serial)

InfiniBand(Serial)

Close integration ofSystem Software andHardware functions

Definable low latencies

In order delivery built into the hardware

Close integration ofSystem Software andHardware functions

Definable low latencies

In order delivery built into the hardware

IP based message passingenvironment

Variable (high) latencies

In-order delivery supportrequired (TCP)

Security and QoS defined

IP based message passingenvironment

Variable (high) latencies

In-order delivery supportrequired (TCP)

Security and QoS defined

Storage Connectivity L Di t


52/75


Long Distance

Long Distance to StorageWhy???

Large campus environment Separate data centers that

need improvedmanageability

Disaster recovery orbusiness continuity

Vendors want todemonstrate unique

features

FC

Short distance


53/75


Storage Connectivity Long Distance


54/75


Long Distance

Before you go the distance, consider - Applications differ in terms of latency and bandwidth

Application read and write characteristics vary some applications read more data than they write

some write in small data blocks, others large

some have serial activity while others are random Where is caching being done?

in the server, array or application

Storage Connectivity Long Distance


55/75


Long Distance

Before you go the distance, consider - Cross-site replication techniques vary considerably*

Should you mirror the data synchronously, asynchronously orsemi-synchronously?

Should the storage array do the replication, or should theoperating system or application do the replication?

Should a virtualization appliance handle the replication?

Should you make full replicas of data or just snap copies?

Check outSNIA Tutorial:

Data Protectionwith StorageNetworks


Data Protectionwith StorageNetworks

Storage ConnectivitySAN Extension Options


56/75


SAN Extension Options

FCIP over public or private IP (Frame Relay or T1/T3)*

FCIP over SONET/SDH

FC directly over SONET/SDH FC over DWDM or CWDMCheck outSNIA Tutorial:

Metropolitan andWide AreaNetworks

Check outSNIA Tutorial:Metropolitan andWide AreaNetworks

FCIP

SONET

Ethernet/PoS

IP

TCP

FCIP

FC

FC over ATM

SONET

ATM

AAL

FC

DWDM

FC over SONETover DWDM or

CWDM

SONET

FC

FC over SONET

SONET

FC


57/75



58/75


59/75

Page 59

Storage Basics

New way to look at Storage

Information Life Cycle Management

(ILM) and Tiers of Storage (TOS)

Information Lifecycle


60/75


Information lifecycle: The creation and/or acquisition of the data information comes

into the organization either by being created by one or moreindividuals or by being acquired through e-mails, faxes, letters,phone calls, etc.

The publication of the data some information needs to be

published, either in print form or on a companys intranet or apublic Web site.

The retention and/or removal of the data some information mustbe archived for later use, and some information has a finite

purpose and can be discarded once it has served its purpose oris no longer valuable to the organization.

http://www.webopedia.com/TERM/I/ILM.html

Information LifecycleMANAGEMENT


61/75


MANAGEMENT

IL Management determining how the information is stored based on

how high of a priority the access of the information hasin the organization at any given moment.

At each stage in the informations lifecycle, the

management infrastructure must determine the bestsoftware, hardware and storage medium required forthe information at that stage, and how those factors

differ as the data move through the lifecycle.Check out

SNIA Tutorial:BusinessContinuityAnd HA


BusinessContinuityAnd HA

Tiers of Storage (TOS)


62/75


Hardware driven*

High-end storage SuperRAID Array

Middle tier Raid Array

Low end storage jbod

features, functionality, capacity, I/Ocapabilities, etc

features, functionality, capacity, I/Ocapabilities, etc

*Moving beyond data storage By Ed FrauenheimStaff Writer, CNET News.com

Tiers of Storage (TOS)


63/75


Policy Driven First

How valuable is this information now? What kind of performance do you need now?

Where is the best place for storing it?

What is the lowest cost that meets your requirements?

Secondly What Information needs to be protected and how?

Backup tiers (Primary, secondary, nearline, online, offline)

What is the value of information over time? Migration of data

Case Study Hardware TOS


64/75


Stanford University Information Technology Systems and Services

Feature High- End Mid-Tier Low-EndUptime >99.999%

< 6 min down/year>99.99%< 1 hour down /year

>99.9%< 9 hours down /year

Connectivity FC, SCSI, ESCON,FICON

FC FC

Ports 16 - 96 4 - 8 2

Internal Disks SCSI or FC

(73GB/146GB)

SCSI or FC

(73GB/146GB)

ATA (180GB/250GB)

RAID levels 0, 1,3,5 0,1,3 0,1

Maximum LUNs 20,000 1024 32

Cache Size 16GB 64GB 2GB 16GB 512MB 1GB

Non-DisruptiveUpgrades

Yes Yes Maybe

Upgrade Cost per GB >$120 < $25 < $7

Architecture Monolithic or Modular Modular Modular


65/75


66/75

Page 66

Storage Basics

Audience Poll and Q&A

Audience Poll


67/75


Are you an end user? Reseller?

Vendor?

Do you have networkedstorage in place today?

NAS or SAN?

Do you plan to within 6 or12 months?

Have you attended otherSNIA Tutorials?

At this meeting or otherSNIA event?

Have you been SNIA-

certified? Do you or your staff plan on

taking exam within 6 to 12months?

Q&A / Feedback


68/75


Please send any questions or comments on thispresentation to SNIA:

[email protected] an y t h a n k s t o t h e f o l lo w i n g i n d i v i d u al s

f o r t h e ir c on t r i b u t i o n s t o t h i s t u t o r i al .

SNIA Education Committee

Elaine Silber Infinity I/O Barry Walker Infinity I/O

Bob Lockhart Neoscale SW Worth Microsoft

Brandy Bartyon Medusa Labs Jim Nelson Vixel

Howie Goldstein - HGAI Sam Samuel Infinity I/O

Leroy Budknik Knowledge Transfer Barbara Craig QLogic

John Moores Sandial Systems Ronnie Koch Infinity I/O, Africa


69/75

Page 69

Storage Basics

Appendix

Hidden Slides

RAID 0Striped Volumes


70/75


Data blocks written sequentially to each disk in turn

Not really RAID No redundant check data

Single disk failure can result in loss of all data Better performance than single disk access for large files

Good where performance is more important than redundancy

Volume n

Volume B

Volume A

SERVER RAID STORAGE

Block 5 Block 6

Block 3 Block 4

Block 1 Block 2

RAID 1Mirrored Volumes


71/75


Data blocks written to both disks at once

100% data redundancy means no data loss

If one disk fails, data can be retrieved from mirrored disk

Requires two disk write operations per block but only one read

Good for small files where security is paramount

Volume n

Volume B

Volume A

SERVER RAID STORAGE

Block 3 Block 3

Block 2 Block 2

Block 1 Block 1=

RAID 0+1Mirrored Array


72/75


Data blocks written to each disk in turn then copied to mirrored array

Combines RAID 0 performance with RAID 1 Redundancy

If one disk fails, the array becomes a RAID 0

Limited scalability and double the cost

Volume n

Volume,me B

Volume A

SERVER RAID STORAGE

Block 5 Block 6

Block 3 Block 4

Block 1 Block 2=

Block 5 Block 6

Block 3 Block 4

Block 1 Block 2

RAID 4Block-Level Stripingwith Parity Disk


73/75


Data blocks written sequentially to each disk in the array

Similar to RAID 3 but generally performs better because data isaccessed in blocks instead of bytes

Volume n

Volume B

Volume A

SERVER RAID STORAGE

Block A3 Block B3

Block A2 Block B2

Block A1 Block B1

Parity 3

Parity 2

Parity 1

Parity

RAID 5Striping withDistributed Parity


74/75


Data blocks written sequentially to each disk in turn

Parity block computed for each row and distributed across all disks

If one disk fails, data can be retrieved using parity blocks

Parity calculation overhead reduces write performance

Good aggregate transfer rate during read

Volume n

Volume,me B

Volume A

SERVER RAID STORAGE

Block A3 Parity 3

Block A2 Block B2

Block A1 Block B1

Block C3

Parity 2

Block C1

Block D3

Block D2

Parity 1

RAID 10Mirrored Striping Array


75/75


Data blocks written sequentially to each mirrored disk array

Combines RAID 0 performance with RAID 1 redundancy

If one disk fails, data can be retrieved from mirrored disk

Parity is not calculated so good write performance

Volume n

Volume,me B

Volume A

SERVER RAID STORAGE

Block 5 Block 5

Block 3 Block 3

Block 1 Block 1=

Block 6 Block 6

Block 4 Block 4

Block 2 Block 2=

Storage for Networking Professionals

Documents

Transcript of Storage for Networking Professionals