1 RAID LEVEL 0 Description

Apresentação em tema: "1 RAID LEVEL 0 Description"— Transcrição da apresentação:

1 www.pontoderedes.bogspot.com 1 RAID LEVEL 0 Description
RAID Level 0 is not, strictly speaking, a RAID array does not provide a mechanism for assuring high data availability. Disk striping. In traditional storage systems, a logical record or a block of data is entirely contained on a single disk drive. This causes the host application to "wait its turn." With RAID Level 0 a block of information or a logical record is broken down into smaller pieces and distributed across multiple disk drives. Utilizes an external controller and processor to control the array, additional performance improvements are achieved over using a host based "software array". Operating During Failure Improved performance is the only enhancement delivered by RAID Level 0. There is no protection that provides for improved data availability. In fact, because data is striped across multiple drives, it is likely that the failure of one drive will live "holes" in the data rendering the entire array unusable. Additionally, the MTBF of the array declines as the number of drives increases. Therefore, RAID Level 0, by itself, is unsuitable for data that cannot easily be reproduced or for mission- critical applications requiring continual data availability. Performance Characteristics The concurrent execution of I/O requests is the primary reason that RAID Level 0 can transfer data faster that a single disk. RAID Level 0 provides high performance for a wide variety of I/O intensive applications. When configured with all drives acting in unison, RAID Level O can provide high data transfer rates of large blocks of data. This makes Level 0 very useful for imaging applications, audio/video processing, sequential file processing, and the storage of program libraries. When configured as an Independent access array with all drives separately processing I/O requests, RAID Level 0 can provide faster response to frequent I/O requests. This makes it useful for transaction processing and general file services for multi-user systems. Using Scatter read and write gathering, the performance of applications that make large I/O requests is substantially improved. If the requested data is larger than one stripe, the array can cause the disk (s) to read/write adjacent stripes during the same I/O request rather than processing separate request for each stripe level. In the RAID Level 0 diagram on page 6 for example, if the application needs the first six blocks (i.e., 0123, 4567, 8901, 2345, 6789, 1234), an array supporting Scatter read would instruct Drive 1 to return 0123 and 6789, Drive 2 to return 4567 and 1234, Drive 3 to return 8901 and Drive 4 to return This uses 4 I/O requests instead of six. The same reduction in I/O requests occurs with write gathering since a write updates multiple adjacent stripes with a single I/O request. 1

2 Níveis Raid RAID Level 0 - Striping
Uso de múltiplos Discos para a formação de um único Disco lógico. Performance na implementação do RAID Alta performance na Escrita e Leitura (Write and Read) performance relacionada com o aumento da quantidade de Discos. Os Discos Rígidos são distribuídos utilizando-se uma tamanho definido de “stripe” durante a configuração Deve ser otimizado em conjunto com o Sistema Operacional para uma performance otimizada As pequenas solicitações que possuem o mesmo tamanho de “stripe” são transmitidas a um único Disco Rígido, as solicitações maiores são divididas e transmitidas a múltiplos Discos Rígidos em paralelo A capacidade é a soma do número de discos no “array” Não proporciona proteção contra falhas de hardware, somente performance.

3 Resumo Nível: RAID 0 Todos os Dados Foram Perdidos
RAID-0, “Striping”, todos os Discos estão disponíveis para dados, Inseguro. Utilizado para armazenamento temporário Qualquer Disco que falhar causa a perda dos dados 10 = Total de Discos 10 = Capacidade Usável 0 = Redundância de Discos 10 = Performance Leitura 10 = Performance Escrita Dados Todos os Dados Foram Perdidos

4 4 RAID Level 1 Disk Mirroring Description
RAID Level 1, or disk mirroring, stores images of the data on each disk drive. I/O requests are processed differently depending on whether they are read or write requests. For read requests, the array has the ability to make a choice as to which disk to read from. This can be done either on a random basis or on an alternating basis resulting in load sharing. During a write operation, both drives in a pair must be updated. A single write request for the host is translated into a write for each drive in the pair. Operating During Failure If either disk in the pair fails, the remaining disk can carry on with its normal activities without any downtime. The reconstruction is usually transparent to the host system. However, performing the reconstruction does entail overhead especially while continuing to provide normal I/O operations. Performance Characteristics RAID Level 1provides improved read performances compared to traditional disk subsystems due to Parallel I/O paths. 4

5 Níveis Raid RAID Level 1 - Espelhamento
O RAID 1 proporciona alto nível de tolerância a falhas Cada solicitação de I/O é espelhada em um segundo Disco Rígido O RAID 1 trabalha com múltiplos de dois Discos Rígidos- o set primário e o set espelhado – dobrando também o custo por GB da solução Proporciona o mesmo throughput durante a escrita Oferece mais performance durante a leitura (read) – A controladora RAID executa o mesmo comando de leitura em cada Disco – assim que um dos discos disponibiliza os dados, ele é automaticamente transferido para o servidor e o próximo comando de leitura pode ser processado. Oferece proteção contra falhas nos discos com taxa de 1 para 1 Assim que um disco falha, automaticamente o espelho assume, porém caso ocorra a falha no espelho não existe mais segurança Pode proteger contra desastre naturais/físicos porém o espelho deve estar instalado remotamente (em outro local físico) acarretando em um custo adicional $$. Não protege contra vírus digital ou acidentes/delete intencional, já que 100% do disco está sincronizado com o outro Implementação de Alto Custo Controladora RAID + Custo por GB de chega a ser o dobro se compararmos com uma unidade simples.

6 Resumo Nível: RAID 1 RAID-1, “Mirroring” rápido, 50% dos Discos estão disponíveis para dados, extremamente seguro. Utilizados para informações importantes Qualquer Disco espelhado pode falhar, e os dados permanecem salvos 2 = Total de Discos 1 = Capacidade Usável 1 = Redundância de Discos 2 = Performance Leitura 1 = Performance Escrita Data Copy Os dados permanecem Salvos Todos os Dados Foram Perdidos

7 RAID Level 3 uses parity information to create data redundancy
RAID Level 3 uses parity information to create data redundancy. Data is distributed across all the disks in such a way that the data can be mathematically reconstructed. Data is striped across the data disks using bit or byte striping so a logical record or block spans across multiple drives. The drives in the array are accessed in parallel: when information is read, it is gathered simultaneously from all the data drives, reassembled into a logical block then passed to the host. A separate disk is used to store parity information about the disk drives. Operating During Failure A disk failure may be one of the data disks or the parity disk. In either case, however, the ability to read or write data is not impaired. If one of the data disks fails during a read operation, information on the parity disk is combined with the data from the remaining drives to mathematically determine what the missing data should be. During a write with a failed data disk, the data is written to the remaining drives. When the failed disk is replaced, all the data that should be on it is rebuilt using the information from the other data drives combined with the parity values. If the parity disk fails, information is retrieved from the data disks as usual. Performance Characteristics Provides excellent I/O performance for applications in which large blocks of sequentially located data are being used, such as image/audio/ video processing, and scientific data collection. The drives are operation in unison and are "tied up" fulfilling an I/O request. In a failed data disk mode, regardless of which data disk has failed, every request for data will require regeneration of the missing data because the block being requested spans multiple drives. Rebuilding the data on a replace data or parity disk involves repeated access to all drives in the array. 7

8 Paridade e ECC (5+10+?+2+7 = 27) Bit Error
Utiliza metodologia de reconstrução dos dados perdidos, adicionando-se informações adicionais, comumente chamada de “overhead” Utiliza cálculos Matemáticos Polinomiais e operações inversas È utilizada uma equação de 5th Grau para encontrar a soma e o resultado. O método de detecção de erros com paridade em computadores funciona apenas para a detecção do erro. (5+10+?+2+7 = 27) Bit Error Método de Paridade no barramento SCSI Binary Number 128 64 32 16 8 4 2 1 Parity Bit Parity Number 6 3 33 Parity Sum Error

9 Níveis Raid RAID Level 3 – Acesso Sincronizado com disco dedicado de paridade
O Uso de mais Discos possibilita um aumento do throughput Um único de Disco de paridade pode ser um problema de gargalo na performance Proteção contra falha de Disco com razão de 1 para vários A Performance é degradada durante o acesso, e especificamente durante a reconstrução em caso de falha (rebuild) de um disco rígido (demorando horas) Não oferece proteção contra desastre físicos Não protege contra vírus digital ou acidentes/delete intencional, já que os dados são protegidos por um esquema de paridade ECC (com esquema e atualização em tempo real) Implementação de Alto Custo Custo da Controladora RAID + Discos Rígidos com sincronização de spindle 1 Disco Rígido adicional para Paridade Implementado como um sistema RAID (gabinete stand alone)

10 RAID Level 5 Distributed Parity Independent Access Arrays Description
Also uses parity information to provide data redundancy. However, data is striped across drives using larger, multi-byte units (unlike RAID Level 3). RAID Level 5 arrays can process multiple requests simultaneously as long as the data being requested is located on different drives. Parity information itself is distributed across all drives just like user data. When a read operation takes place, the array management software determines which stripe(s) in which drives(s) will be required and then issues the request to the appropriate drive (s). During a write operation, the array calculates the parity and issues commands to write the new data to the appropriate drive(s). When a write occurs in a RAID Level 3 array, the completely updated stripe is written to all drives within the stripe. The newly calculated parity is written to the parity drive at the same time. In a RAID Level 5 array, a record update may only involve data on one of the drives. The others drives do not need to be involved and should be free to process other requests. However, in order to calculate the parity of the entire stripe, the array needs to know the contents of the entire stripe. Operation During Failure When writing entails updating the parity on a failed disk, information is written to the other disk(s) as normal but parity information cannot be stored. When the failed disk is replaced, it must be brought up to date. When attempting a read operation that involves a failed disk, information from its corresponding parity (stored on a different disk) is combined with data from the remaining drives to mathematically determine what the missing data should be. During a write to an array with a failed data disk, if the write request only involves still-functioning drives, the read-modify-write operation takes place as normal: there is no need to involve the failed drive. If the write request includes storing data on a failed drive, the data of course, cannot be written to it. However, the parity value necessary to allow reconstruction of the data can be written because the parity will be stored on one of the other drives that is still functioning. Performance Characteristics Well suited for applications involving a large number of concurrent read request where high data access rates are desirable and there is a higher proportion of reads compared to writes. Cache memory and other data management techniques are designed to minimize the write penalty. When a read or write is requested to a failed disk, all drives must be accessed to gather the remaining data and parity in order to calculate the missing information. 10

11 Níveis Raid RAID Level 5 – Acesso Independente com paridade distribuída
O uso de mais discos possibilita o aumento da performance RAID 5, é geralmente utilizado em ambientes OLTP Proteção contra falha de Disco com razão de 1 para vários Qualquer disco rígido (1) pode falhar e mesmo assim o sistema permanece intacto Não oferece proteção contra desastre físicos Não protege contra vírus digital ou acidentes/delete intencional, já que os dados são protegidos por um esquema de paridade ECC (com esquema e atualização em tempo real) Implementação de Alto Custo Capacidade com perda pela Paridade, o crescimento dos dados aumenta o custo por GB Custo da Controladora RAID + Implementado como um sistema RAID (gabinete stand alone)

12 Resumo Nível: RAID 5 RAID-5, rápido, todos menos um disco está disponível para dados, Segurança. Utilizado com Armazenamento Primário Caso um Disco falhe ”OK”, O segundo Disco ocasiona a perda total dos dados! 10 = Total de Discos 8 = Capacidade Usável 1 = Redundância de Disco 8 = Performance Leitura 8 = Performance Escrita Data Paridade Hot Spare Os dados permanecem Salvos Todos os Dados Foram Perdidos

13 13 RAID Level 0+1 Striped Mirrored Arrays Description
RAID Level 0+1, also known as RAID 10, combines the data availability advantages of RAID Level 1 disk mirroring with the performance benefits of RAID Level O disk striping in a single array. A logical record or block is broken down into smaller pieces. Each piece is distributed across different disk drives (like RAID Level 0) and each of these disk drives has a mirrored twin (like RAID Level 1). With load balanced striping, the Customer file is distributed across multiple drives. The RAID Level 0+1 array also provides high performance for application that need high data transfer rates (i.e. read/write request or large blocks of data). This architecture is very reliable since a duplicate copy of all data is available even in the highly unlikely event of a simultaneous failure of more than one disk. Operating During Failure If either disk in the pair fails, the remaining disk can carry on with its read and write activity without any downtime. The array hardware and software ensure that the switch-over ensure that the switch over to the "backup' drive is transparent to the host. Performance Characteristics The high I/O performance of RAID Level 0+1 arrays comes from two levels of load balancing: Distributing the data across multiple drives. Provides a secondary level of load balancing by sending the request to the least busy disk to complete the request more quickly. Provides high performance for a variety of I/O-intensive applications. 13

14 Combinação de níveis RAID
Múltiplos níveis RAID 0, 1, 3, ou 5 Arrays podem ser “striped” entre eles para a criação dos níveis RAID 00, 10, 30 ou 50 Proporciona mais capacidade Frequentemente chamado de composição RAID Proporciona um melhor “load balancing”. Interface da Aplicação RAID 0 Stripping RAID 1 ARRAY

15 Resumo Nível: RAID 0+1 ------------------------
O RAID-10, é extremamente rápido, metade dos Discos estão disponíveis para Disco/ dobro do Preço, Seguro. Usado para dados primários e críticos Caso um Disco falhe ”OK”, O segundo ocasiona a perda dos dados! Os dados permanecem Salvos Todos os Dados Foram Perdidos 20 = Total de Discos 10 = Capacidade Usável *2 = Redundância de Discos 20 = Performance Leitura 10 = Performance Escrita Data Grupo #1 Dados Críticos Grupo #2 Espelho (Cópia)

16 Os dados permanecem Salvos
RAID 5+1 oferece 4 discos de capacidade porém com necessidade total de 10 discos: Dados Paridade Grupo #1 Dados Críticos Grupo #2 Espelho (Cópia) Striping Paridade Os dados permanecem Salvos Todos os Dados Foram Perdidos 16

17