講稿系統(tǒng)類-ibmaixdsoracle帶庫chitecture

Jun,

2011IBM

System

Storage

TS7650G:Introduction

and

ArchitectureAnnouncing:

IBM

acquires

DiligentPriva y

held

CompanyFoundedJune2002Headquartered

in

MassachusettsDevelopment

in

IsraelMore

than

100

employeesInstallations

in

100

Fortune

500

CompaniesThis

acquisition

closed

on

March

11,

2008.

Diligent

is

now

a

subsidiary

ofIBM.“…ProtecTIER

represents

a

breakthrough

that

will

enable

enterprisecustomers

to

fundamentally

alter

their

use

of

disk

for

data

protectionand

archiving.”GlassHouse

Technologies

–

July

2006Virtual

Tape

LibraryA

virtual

tape

library

(VTL)

is

a

data

storage

virtualization

technology

used

typicallyfor

backup

and

recovery

purposes.

A

VTL

presents

a

storage

component

(usuallyhard

disk

storage)

as

tape

libraries

or

tape

drives

for

use

with

existing

backupsoftware.Virtualizing

the

disk

storage

as

tape

allows

integration

of

VTLs

withexisting

backup

software

and

existing

backup

and

recovery

processes

and

policies.The

benefits

of

such

virtualization

include

storage

consolidation

and

faster

datarestore

processes.In

some

cases,

the

data

stored

on

the

VTL's

disk

array

is

exported

to

othermedia,such

as

physical

tapes,

for

disaster

recovery

purposes

(scheme

called

disk-to-disk-to-tape,orD2D2T).CABCAABBACABCAABBACA

A

B

A

C

BB

AWhat

is

DataDeduplicationData

deduplication

is

a

technology

used

to

reduce

the

amount

ofspacerequired

to

store

data

on

disk.

This

is

achieved

by

storing

a

singlecopy

of

data

that

is

backed

up

repetitively.One

copy

of

each

element

is

retained,

pointers

are

created

for

the

duplicateitems,

and

the

duplicate

items

are

not

stored.Deduplication

Methods1.

Hash

Based?Creates

a

Hash

for

a

slice

of

data(

哈希值大小128bits

或160bits)Deduplicates

when

the

same

Hash

is

computed

on

new

data哈希

，容易造成數(shù)據(jù)丟失??2.

Content

Aware?Objects

compared

with

the

same

properties,such

as

file

namesPerformsa

bit-to-bit

comparison必須了解數(shù)據(jù)結(jié)構(gòu)，記錄millions

of

file

names??3.

HyperFactor?Compares

new

data

with

previous

data

that

is

similar

in

memoryRequires

a

computational

compare

step

，二進(jìn)制比較，記錄by的數(shù)據(jù)改變1PB

的數(shù)據(jù)索引只有4GB

，可以完全存放在內(nèi)存中evel??5.Reference

Hash

previously

stored

and

save

newHash

and

dataHash

Based

ApproachSlice

next

data

into

chunks

and

look

for

Hash

MatchA

B

C

D

FGenerate

Hash

per

chunkSlice

data

into

chunks

(fixed

or

variable)A

B

C

DGenerate

Hash

per

chunk

and

saveEAh

Bh

Ch

Dh

EhAh

Bh

Ch

Dh

FhContent

Aware

Approach1. Look

through

data

for

fully

qualified

file

namesFile

DFile

A File

B File

C2. Locate

previous

version

offileFile

A

→

extents

1,2,3,4Compare

reference

with

versionFile

A File

B File

CUpdate

reference

to

point

to

new

referenceFile

DHyperFactor

ApproachLook

through

data

for

similarityNew

Data

StreamRead

elements

that

are

most

similarDiff

reference

with

version

–

will

use

multipleelementsElement

A Element

B Element

CMatches

factored

out

–

unique

data

added

to

repository#1#2Post

ProcessingAs

data

is

received

by

the device

itis?temporarily

stored

on

disk

storageData

is

subsequently

read

back

in

to

be

processed

by

a

de-duplication

engineBackup

window

is

not

impactedIncrease

I/O

operations

to

storage

devicesNeed

more

disk

capacityTwo

data

deduplication

processing

methodsInlineAs

data

is

received

by

thede-duplicated

in

real

timedevice

it

isnot

temporarily

stored

on

diskData

written

to

the

disk

storage

isde-duplicated

，

requires

less

disk

capacity8:00

AM2:00

AM8:00

PM8:00

PMPost

ProcessingDe-DupeOverlapVTLTape

LibraryTruckInline

ProcessingBackup

ServerProtecTIER

VTTape

LibraryTruckDe-DupeSLA

isMet8:00

PM2:00

AM

8:00

AM 8:00

PM與后處理相比，

處理方式從備份數(shù)據(jù)進(jìn)入時(shí)就開始進(jìn)行重復(fù)數(shù)據(jù)刪除，能夠縮短備份全過程所需的時(shí)間，更好的滿足用戶

SLA

（服務(wù)級別協(xié)議）的要求Components

of

Data

Deduplication

System1.

Server?With

an

operating

system,

on

which

the

data

deduplication

softwareruns.2.

Data

Deduplication

software?Performs

the

data

deduplication

process3.

Disk

Array?Stores

data

processedIntergrated

or

separa

y?Benefits

of

Data

Deduplication?數(shù)據(jù)備份節(jié)省

空間，相同的物理空間可以節(jié)省成本（磁盤、能耗），減少資源浪費(fèi)?TS7650G

Gateway

Model

：

3958-DD42010

年12

月3

日GA4U

system

x3850

x5

type

7145

server4

個(gè)

8

核

in Xeon

X7650,

2.27Ghz8

根內(nèi)存條，共64Gcache兩塊內(nèi)置300G

SAS

硬盤，raid

1

mirror兩塊雙端口Qlogic

QLE2562

8Gbps

光纖卡Redundant

power

supplyRedundant

fans1

塊四端口1G

網(wǎng)卡S D/DVD

ROM配置VTL

：兩塊雙端口Emulex

LPe12002

8Gbps

光纖卡配置OST

：兩塊四端口1G

網(wǎng)卡Rear

view

of

TS7650G

with

VTL

configurationRear

view

of

TS7650G

with

OST

configurationProtecTIER

overviewSoftware

solution

that

resides

on

standard

Linux

server(Red

Hat

5.4)Emulates

a

tape

library

unit,

includingdrives,

cartridges

and

roboticsUses

FC-attached

disk

array

as

the

backup

mediumBackup

Server

FC

Disk

ArrayProtecTIER

VTApplicationProtecTIERServerVirtual

Tape

Library“Filtered”

dataFC

SwitchTS7650GBackupServersDiskArraysRatio

of

Repository

to

Index:

250,000

:

1New

Data

StreamRepositoryMemoryResidentIndexProtecTIER?

with

HyperFactorTMHyperFactorTMAlgorithm

filters

redundant

dataHigh

speed

pattern

matching1PB

of

physical

capacity

(User

Data)I/O

throughputup

to

1000

MB/sperclusterHYPERFACTORNEWDATANEWDATANEWDATANEWDATANEWDATANEWDATABackup

application

writes

data

toProtecTIER

as

it

would

to

tapeData

goes

through

HyperFactordeduplicationengineOnly

unique

data

is

storedExisting

duplicate

data

is

referencedWhen

a

data

object

expires

or

isoverwritten,

references

are

removedFree

space

is

reclaimed

and

reused12345EFGHIJA

B

C

DOverview

of

ProtecTIER

OperationsDisk

ArrayProtecTIER

ServerBackup

Application

ServerFCFCTCP/IPProtecTIERManagerTS7650GatewayProtecTIER?

ComponentsApplianceProtecTIER?

Conceptual

Flow

ChartLocatepattern(RAM

search)RepositorydataNew

backupdataRead

similarpattern

fromrepositoryCompute“delta”Storenew

backup“delta”DiligentcompressionMemoryindex2:1Virtual

Tape

ConceptHyperFactor

Data

StorageProtecTIER

RepositoryRAID-10RAID-5ProtecTier

EE

v2.5

Base

SoftwareProtecTier

Enterprise

Edition

v2.5

supports:Inline

data

deduplication

powered

by

HyperFactor

technologyUp

to

16

virtual

tape

libraries

per

clusterUp

to

512

virtual

tape

drives

per

clusterUp

to

512,000

virtual

cartridges

per

clusterUp

to

1000

MB/s

or

more

sustained

inline

deduplication

backup

performanceScales

to

1

PB

of

physical

storage

over

25

PB

of

userdataEmulation

of

IBM

TS3500

tape

library

with

Ultrium

3

tape

drivesEmulation

of

Quantum

P3000

tape

library

with

DLT

7000

drivesEmulation

of

DTC

VTF

0100

virtual

tape

librariesMany-to-One

replicationProtecTier

EE

v3.1

Base

Software---latest

releaseProtecTier

Enterprise

Edition

v3.1

provides

the

following

core

differentiators:High

speed

backups:

Up

to

2000

MB/s

or

more

sustained

inline

deduplication

backupperformanceEven

faster

restores:

Up

to

2800

MB/s

or

more

sustained

recovery

performanceReplication:

Support

Many-to-Many

&

Bi-directional

replication

between

two

to

fourVTLProtecTier

systemsProtecTIER

ServerTS7650G

Control

Path

Failover

SupportA

single

virtual

SCSI

mediumchanger

(SMC

or

accessor)

canbe

configured

on

multipleports.Multiple

host

HBA

ports

can

beconnected

to

the

same

virtualSMC

through

differentports.When

a

host

HBA

port

goesdown

or

a port

goes

down,CPF

will

failover

to

thenextavailable

path

to

access

thevirtual

accessor,

withoutinterrupting

the

job

in

progress.TargetPortTargetPortTargetPortPortHBA

PortHBA

PortHOSTCPFProtecTIERServerTS7650G

Control

Path

Failover

in

a

SAN

EnvironmentSANHBA

PortHBA

PortThrough

a

switch,

more

paths

canbe

configured

on

host

to

accessthe

tape

library.When

twopaths

go

down,

CPFcanfailover

to

a

3rd

path,

if

available,

toaccess

the

tape

library

withoutinterrupting

the

job

in

progress.HOSTCPFTargetPortTargetPo

rtTargetPortPortProtecTIERServerProtecTIERServerHOSTCPFUnavailableActiveVirtualDrivesUnavailableAvailableALL

PATHSTO

ANODE

LOSTWITH

CPFVirtualAccessorTS7650G

Control

Path

Failover

in

Dual-Node

ConfigurationClusterTS7650G

DataPath

FailoverSupportEach port

can

configuremultiple

virtual

tape

drives,but

a

given

virtual

tapedrivecan

only

be

configured

on

oneport.Multiple

host

HBA

ports

canbe

connected

to

the

samevirtual

tape

drive

through

aswitch.When

one

of

the

host

HBAport

connection

goes

down,DPF

willfailover

to

the

nextavailable

host

HBA

port

toaccess

the

tape

drive,

withoutinterruptingthe

databackup/restore.TargetPortTargetPortTargetPortPortProtecTIERServerDrive

1SANHBA

PortHBA

PortHOSTDPF30Backup

ApplicationServerSingle

Virtual

LibraryRepositoryTS7650

clusterServer

131Server

2TS7650G

Cluster

withSingle

Virtual

LibraryServer

1

Failure:Lose

Control

PathLose

access

to

libraryCan

manually

reconfigure

controlpath

toServer

2Server

2

FailureContinue

with

3

drives

on

Server

1Assign

Control

Pathto

only

one

server

in

clusterBackup

ApplicationServerVirtual

Library

ARepositoryTS7650

clusterVirtual

Library

BServer

132Server

2TS7650G

Cluster

withTwo

Virtual

LibrariesEither

Server

FailsLibrary

on

remaining

serveravailableUsethree drives

withbackup

app

(opt)Lose

access

tocartridges

inlibraryattachedtofailed

serverCan

manually

reconfigurecontrolpath

to

remaining

serverAssign

Control

Pathto

only

one

server

in

clusterBackup

Application

ServerCPF

EnabledSingle

Virtual

LibraryR