The reliability of storage devices, today as in the past, is a fundamental requirement in those areas where the unavailability of information or, even worse, their complete loss mean interruption of services and substantial losses money. Thanks to the rapid technological development the hard drives have become increasingly reliable and fast, but it is still the danger of mechanical failure, electronic and those resulting from wear or damage magnetic media.
While reliability can be improved with particular care in the design and implementation phases the product, the wise use of multiple units allows for data reliability and fault tolerance far superior to that offered by the best record in the market, the technology Raid is rooted precisely on this idea. The term Raid stands for Redundant Arrays of Inexpensive Disks (redundant array of inexpensive disks) was used for the first time by researchers at the University of California at Berkeley in 1987.
Over the years, have been developed in many configurations, each identified by a number following the word raid, of which only some have proved successful for the market. Before presenting the three configurations più diffuse in ambito workstation e server riteniamo necessario familiarizzare con alcuni concetti che definiscono l’organizzazione gerarchica di un sistema Raid: Disco Fisico, Array Fisico, Array Logico e Disco Logico. Il disco fisico è l’unità singola vera e propria che funge da mattone elementare nella realizzazione della configurazione Raid.
L’insieme organizzato di uno o più dischi fisici costituisce l’array fisico; le configurazioni Raid più semplici utilizzano un solo array fisico di dischi, mentre quelle più complesse possono appoggiarsi a due o più array fisici distinti. L’array logico è il primo livello di astrazione dall’hardware ed è il frazionamento o la combinazione one or more physical arrays, typically a logical array corresponds to a physical array, but you can, for example, that a logical array is composed of several physical arrays.
logical disks, finally, are obtained from the array and logical disk volumes to correspond to the operating system. At this point it is easy to see that a logical disk can hide complex management mechanisms that remain invisible to both the operating system is delegated to the user as appropriate controller.
Raid 0 (Striping)
The Raid 0, which is enjoying particular success in the desktop, is to configure multiple drives in parallel in order to increase performance by writing and reading simultaneously on all units. If this configuration is used as a single improper use of RAID as affect the principle of reliability: not providing any type of parity checking on the data, the failure of a unit that makes up the array results in loss of all information . Since the increase of the number of disks increases statistically the probability of any fault, it follows that, because of the lack of parity, a Raid 0 configuration , while increasing performance, reliability leads to a decrease proportional to number of disks che la costituiscono.
Una caratteristica delle configurazioni striping è la configurazione delle strisce (stripes), ovvero la scelta della dimensione dei blocchi nei quali vengono spezzate le informazioni. La scelta di questo parametro è figlia del numero di dischi impiegati nell’array e del tipo di software impiegato, ovvero se si lavora su file di piccole, grosse o medie dimensioni. Raid 1 (Mirroring e Disk Duplexing)
Il Raid 1 costituisce l’implementazione più semplice della protezione dei dati e consiste di due dischi o array configurati in modo che uno sia la copia speculare e in tempo reale dell’altro. Di base non si riscontra alcun incremento di prestazioni in quanto a branch of the configuration operates in backup mode, but there are controllers evolved to optimize stages of reading through the use of both arms.
depending on the architecture used to build the array we talk about how Disk Mirroring or Duplexing: the first indicates the configuration of two in a RAID 1 array managed by a single controller, while the second refers to a configuration consisting of two controllers, each of which is entrusted with the management of a branch of the configuration.
Raid 5 Raid 5 mode belongs to the family of configurations of type striping with parity: the information is divided on different drives in the array in a similar way to a solution Raid 0, but on all the disks are stored blocks to control the parity. This device allows, in case of failure of one of the units, rebuilding on the fly data using information from the remaining disks and the code of parity. We sense that in case of failure of the quality controller, specifically the processor and the amount of cache, affecting performance.
The Raid controller also provides information on the reconstruction of the damaged or replaced by hard disk hot spare, or an empty unit on which they are regenerated the information of the failed disk on hold che questo venga sostituito. A differenza delle modalità 0 e 1 diffuse anche in ambito desktop grazie all’integrazione di controller specifici su molte schede madri di fascia medio-alta, le configurazioni Raid 5 sono ancora appannaggio dei sistemi workstation e server. Per implementare una configurazione Raid 5 efficiente è infatti necessario un controller di tipo evoluto, ovvero dotato di un chip dedicato all’elaborazione del codice di parità e una memoria buffer che ha lo scopo di migliorare le prestazioni.
