ahorro energético archivado de backups
TRANSCRIPT
Ahorro energético en el
archivado de backups.archivado de backups.
Miquel MorellMiquel Morell
SM&C ConsultorsSM&C Consultors
•• "If you want to save power, shut everything off "If you want to save power, shut everything off
and and put it on tapeput it on tape. But if you have data that's . But if you have data that's
time sensitive and performance sensitive, time sensitive and performance sensitive,
that's a different story.“that's a different story.“that's a different story.“that's a different story.“
Greg Schulz, founder of The StorageIO GroupGreg Schulz, founder of The StorageIO Group
Consumo en un datacenter
Power (Watts)Power (Watts)Power (Watts)Power (Watts)
37%37%37%37%
23%23%23%23%40%40%40%40%
serversserversserversservers storagestoragestoragestorage restrestrestrest
Kevin Kettler, CTO Kevin Kettler, CTO
Dell Inc. Dell Inc.
Storage-specific power-saving methods
•• Reducing the amount of data through Reducing the amount of data through archivingarchiving, , data data
deduplicationdeduplication and and compressioncompression
•• Tiering storageTiering storage by using more powerby using more power--friendly media, friendly media,
such as solidsuch as solid--state disk, newer highstate disk, newer high--capacity drives capacity drives
that store more data without using extra power and that store more data without using extra power and that store more data without using extra power and that store more data without using extra power and
even tape even tape
•• Improving storage management though Improving storage management though thin thin
provisioningprovisioning and infrastructure resource management and infrastructure resource management
•• Using Using storage virtualizationstorage virtualization to consolidate storage to consolidate storage
resources the way server virtualization helps resources the way server virtualization helps
consolidate server. consolidate server.
Greg Schulz, founder of The StorageIO GroupGreg Schulz, founder of The StorageIO Group
1 TB Data on Different Drives
6,096 kWh/yr
94%
87%
15K 73 GB 15K 146 GB 10K 300 GB 7.2K 500 GB
787 kWh/yr1,434 kWh/yr
3,048 kWh/yr
73%
7.2K 750 GBSATA II
525 kWh/yr
50%
7.2K 1 TBSATA II
393 kWh/yr
25%
High Capacity Disks Consume Menos Energy
SIN Thin provisioning = Tradicional
Database = 1TbLUN = 1Tb
Disco = 1.3 Tb
Datos = 300Gb
100010111010010101010101010101000101110101001010010101010001011101010010100100101010100101010100000111000000100100010001000100100101001010011111111000101101010001011101010010100101101000101110101001010010000000100100010001000100101000101110100101010101010110001011101001010101010101001011101010001011101010001010100101001010100101000010101000101110101001010010000000100100010001000100101000101110100101010101010100101110101000101110101000101010010100101010010100001010100010111010100101001010001011101001010101010101101000101110101001010010101010001011101010010100101010101001010101000001110000001001000100010001001001010010100111111110001011010100010111101001010010110100010111010100101001000100010001000100101000101110100101010101010100101110101000101110101000101010010100101010010100001010100010111010100101001010
1Tb 1Tb
CON Thin provisioning
Database = 1TbLUN = 1Tb
Disco = 300GbvLUN = flexible
Datos = 300Gb
100010111010010101010101010101000101110101001010010101010001011101010010100100101010100101010100000111000000100100010001000100100101001010011111111000101101010001011101010010100101101000101110101001010010000000100100010001000100101000101110100101010101010110001011101001010101010101001011101010001011101010001010100101001010100101000010101000101110101001010010000000100100010001000100101000101110100101010101010100101110101000101110101000101010010100101010010100001010100010111010100101001010001011101001010101010101101000101110101001010010101010001011101010010100101010101001010101000001110000001001000100010001001001010010100111111110001011010100010111101001010010110100010111010100101001000100010001000100101000101110100101010101010100101110101000101110101000101010010100101010010100001010100010111010100101001010
1Tb 300Gb
Mirroredstorage
WAN
Método tradicional:
Backup-a-Disco y Staging-a-cinta
Clients Server Primarystorage
Backup/mediaserver
Tape library Offsite StorageStagingDisk
StagingRetention/
Restore/Cloning DR ArchiveBackup
Disk Staging Benefits:Disk Staging Benefits:
Keep up with backup Keep up with backup windowswindows
Optimize tape automationOptimize tape automation
More reliable backupMore reliable backup
Some restores from diskSome restores from disk
Critical Enabler: Global CompressionTM
=
Traditional Storage
Enterprise Protection Storage
using Deduplication
=
Deduplicación… pero si es muy fácil….
Vista desde Software Backup : Como un file system o una VTL
Primer Full BackupPrimer Full Backup Incr 1Incr 1 Incr 2Incr 2 Segundo Full BackupSegundo Full Backup
A B C D A E F G A B H A E I B J C D E F G H
Data
Stre
am
A B C D E F G H I JDD OS Storage:Redundancies pooled, compressed
= Unique variable segments (4KB-12KB)
= Redundant data segments
= Compressed unique segments
A B C D A E F G A B H A E I B J C D E F G H
DD OS Global Compression
DD560: 400 GB/hour w/ single-core Xeon
Impacto de la reducción de volumen a lo
largo del tiempo
••33--4x 4x
First full backupFirst full backup
••Reduction FactorReduction Factor
••66--7x 7x FileFile--level incrementalslevel incrementals
••5050--60x 60x Subsequent full backupsSubsequent full backups
••20x 20x Aggregate weekly fulls, daily Aggregate weekly fulls, daily incrementals incrementals
Ventajas de comprimir los backups 20:1
o más !!!
•• Repositorios de Restore Repositorios de Restore
•• Tenga varios meses de Tenga varios meses de
backups online y no solo backups online y no solo
algunos dias !!!.algunos dias !!!.
•• Todas las recuperaciones se Todas las recuperaciones se
hacen de disco.hacen de disco.hacen de disco.hacen de disco.
•• Centros de respaldo con Centros de respaldo con
costes muy asequibles.costes muy asequibles.
•• Solo hay que replicar los Solo hay que replicar los
datos no deduplicadosdatos no deduplicados
•• Se puede replicar utilizando Se puede replicar utilizando
lineas IP ya existentes. lineas IP ya existentes.
•• Se eliminan los proceso Se eliminan los proceso
manuales de manipulación y manuales de manipulación y
translado de cintas.translado de cintas.
La linea de sistemas de Deduplicacion
inline más escalable
Appliance &
Gateway
Series
DD580/DD580g
DD565New
New
New
DD510
DD530
DDX Array Series
Replicator & VTL software options
New
DD510DD510 DD530DD530 DD565DD565 DD580/gDD580/g DDXDDX
Speed: GB/hourSpeed: GB/hour 290290 360360 630630 800800 12.8 TB/hr12.8 TB/hr
Logical CapacityLogical Capacity 2525--65 TB65 TB 5555--140 TB140 TB 400400--980 TB980 TB 550 TB550 TB--1.25 PB1.25 PB 8.88.8--20 PB20 PB
Physical CapacityPhysical Capacity 2.25 TB2.25 TB 4.5 TB4.5 TB 7.57.5--23.5 TB23.5 TB 7.57.5--31.5 TB31.5 TB 120120--504 TB504 TB
Up to 16 ControllersInternal or External Storage
Interoperabilidad con el Backup existente
•• DD se ve como un dispositivo NASDD se ve como un dispositivo NAS
•• Se ve como disco (CIFS o NFS)Se ve como disco (CIFS o NFS)
•• Gestión a taves de los comando Gestión a taves de los comando habituales:habituales:
•• Restore File xyzRestore File xyz
•• Clone Disk to TapeClone Disk to Tape•• Clone Disk to TapeClone Disk to Tape
•• Delete File xyz (at end of retention period)Delete File xyz (at end of retention period)
•• Las copias en cinta estan en formato Las copias en cinta estan en formato nativo. (el de siempre !!)nativo. (el de siempre !!)
•• Soporta Synthetic Fulls.Soporta Synthetic Fulls.
•• Legato Networker DBO LicensingLegato Networker DBO Licensing
•• Based on physical (1,2 TB), not virtual Based on physical (1,2 TB), not virtual (25TB)(25TB)
•• Es completamente transparente a la Es completamente transparente a la infraestructura de backup existente!!infraestructura de backup existente!!
Más barato que una VTL !!
Costes licencias....
VTLVTL DataDomainDataDomain AhorroAhorro
NetworkerNetworker
“legato”“legato”
Se licencia por Se licencia por
capacidadcapacidad
Se licencia por Se licencia por
capacidad nativa!capacidad nativa!
~~20:120:1
NetbackupNetbackup Se licencia por Se licencia por GratuitoGratuito 100%100%NetbackupNetbackup
“veritas”“veritas”
Se licencia por Se licencia por
capacidadcapacidad
GratuitoGratuito 100%100%
HP dataHP data--
protectorprotector
Se licencia por Se licencia por
volumen de volumen de
backupbackup
Se licencia por Se licencia por
volumen de volumen de
backupbackup
0%0%
CommVault CommVault Se licencia por Se licencia por
capacidadcapacidad
Se licencia por Se licencia por
capacidad nativa!capacidad nativa!
~~ 20:120:1
Permitase el lujo de tener Disaster Recovery !!
(via Replicación)
Origen Replica
1 TB
20 GB
write
20 GB
•• Los datos se envian al centro de DR Los datos se envian al centro de DR
usando la función de replicación.usando la función de replicación.
•• Solo se envian nuevos “bit segments”, Solo se envian nuevos “bit segments”,
y metadatosy metadatos
• Backup Completo• Gestionado por su Software de Backup
•• Selective by Data Set (vs. System to System)Selective by Data Set (vs. System to System)
•• BiBi--directional (Backup & Replicate to same node)directional (Backup & Replicate to same node)
vtcDir A Oracle Dir S vtcDir A Dir S Oracle
WAN
Flexibilidad en la replica
Source Destination
•• ManyMany--toto--1, Remote Office Data Protection1, Remote Office Data Protection
WAN
Remote Office 1 – DD410
Remote Office 2 – DD430
Dir 1
Dir 2
Dir 3Dir 1Dir 2Dir 3Dir A
Remote Office 3 – DD430
Data Center Hub – DD460
Cross-site Deduplication at
Replica
Dir C Oracle /Home Dir B /Home OracleWAN
Ecológico y asequible.
100 TB
100 TB - Menos potencia electrica100 TB - Menos potencia electrica
- Menos refrigeración
- Menos U de rack
- Menos costes de manmto.
- Menos gestión
- Menos hardware errors
250.000 – 400.000 Euro
75.000 - 80.000 Euro
TCO 5 años (Software Company)
$800,000
$1,000,000
$1,200,000
Maintenance
$-
$200,000
$400,000
$600,000
Current Tape ATA RAID +
Tape
Data Domain +
Tape
Offsite Services
Personnel
Systems
Resumen de las ventajas que aporta
DataDomain
•• Tenga meses de backup Tenga meses de backup a disco y no dias !!a disco y no dias !!
•• Fiabilidad en el RestoreFiabilidad en el Restore
•• Permitase el lujo de Permitase el lujo de tener DR !tener DR !tener DR !tener DR !
•• Facilidad de gestiónFacilidad de gestión
•• “Green” Solution“Green” Solution
•• Económico (<Económico (<11€€/GB)/GB)
•• Compatible con su Compatible con su infrastructura de backupinfrastructura de backup
•• Cutting power consumption through data deduplicationCutting power consumption through data deduplication
•• Bob Dixon, chief architect at U.S. Army headquarters at the Pentagon, found a Bob Dixon, chief architect at U.S. Army headquarters at the Pentagon, found a green benefit in reducing the number of his tape libraries. Dixon said that using green benefit in reducing the number of his tape libraries. Dixon said that using data deduplication appliances from Data Domain Inc. helped cut power data deduplication appliances from Data Domain Inc. helped cut power consumption for the Army's data center. But reducing power consumption wasn't consumption for the Army's data center. But reducing power consumption wasn't the deciding factor in using the data deduplication technology, he said.the deciding factor in using the data deduplication technology, he said.
•• "Our footprint is one"Our footprint is one--tenth the size of the old tape library systems we used to use tenth the size of the old tape library systems we used to use and replaced [because of data deduplication]," he said. "I don't know if it's a big and replaced [because of data deduplication]," he said. "I don't know if it's a big issue, but we want to use our space, power and cooling capabilities wisely."issue, but we want to use our space, power and cooling capabilities wisely."
•• According to the Environmental Protection Agency, the amount of energy According to the Environmental Protection Agency, the amount of energy consumed by data centers doubled between 2000 and 2006. The EPA also consumed by data centers doubled between 2000 and 2006. The EPA also forecasts that power failures and brownouts will affect more than 90% of the data forecasts that power failures and brownouts will affect more than 90% of the data centers in the U.S., and half of large data centers will lack the power and cooling centers in the U.S., and half of large data centers will lack the power and cooling capabilities to run highcapabilities to run high--density equipment in 2008.density equipment in 2008.
•• Nearly every data center can do things to help green its storage networks, Woo Nearly every data center can do things to help green its storage networks, Woo said. "There's no one solution to being green in the storage space," he added. said. "There's no one solution to being green in the storage space," he added. "Different industries will have different requirements that dictate data retention "Different industries will have different requirements that dictate data retention policies. Even departments within organizations have different requirements." policies. Even departments within organizations have different requirements."
10001011101001010101010101 010100010111010100101001010101000101110101001010010 010101010010101010000011100000010010001000100010010 01010010100111111110001011010100010111010100101001011010001011101010010100100000001001000100010001001010001011101001010101010101100010111010010101010101010010111010100010111010100010101001010010101001010000101010001011101010010100100000001001000100010001001010001011101001010101010101 00101110101000101110101000101010010100101010010100001010100010111010100101001010001011101001010101010101 010100010111010100101001010101000101110101001010010 010101010010101010000011100000010010001000100010010 01010010100111111110001011010100010111101001010010110100010111010100101001000100010001000100101000101110100101010101010100101110101000101110101000101010010100101010010100001010100010111010100101001010