Disk arrays and levels of RAID
We have already dealt with the
basic principles of organization of computer memory. Here more detail the
problems of memory, because it is a major long-term data repository.
As for RAM, for the most important characteristics of memory are speed data
exchange and preservation. Obviously, because of the large size of memory loss
of data stored in it because of equipment failure or refusal is a more serious
problem than information loss of memory, and may require significantly more time
for recovery processes calculations. Therefore, to the disk facing greater
demands on reliability. One way to improve the reliability of memory is the
oversupply of disk array to restore the original data, with failures and denials.
One of the technologies improve the reliability of memory was called RAID -
Redundant Array of Inexpensive Disks - redundant array of inexpensive disks.
RAID technology is based on three main methods of recording and protecting
distribution segments of the sequence data on disks with a certain cyclical
sequence, like discussed above for RAM; *
mirror image discs; *
computation of amounts.
Consecutive placement allows you to create a large amount of memory volume and
accelerate the execution of transactions recording and reading data, as the
first segment involves recording data on the first disc, the second - for the
second and so on. In this case array productivity increases, because the
processor or controller input output starts to burn regular segment data for the
next drive before the recording has finished the previous segment. Further gains
in productivity achieved connecting different groups of disks to individual
Mirror display storage and computation of control amounts lead to the emergence
of redundant information that ensures the restoration of lost because of
equipment failure or refusal of data.
Various schemes implementation of disk arrays levels were called RAID.
Level 0 RAID system is not a fault-and quite often such an organization of
memory does not belong to the RAID array. In RAID 0 systems perform only
alternating segments posting data on disks. They apply when the increased
reliability of data storage is not very important, because failure of one drive
led to the loss of the entire stored in an array of information. Such an
organization of memory is used, for example, for tasks editing of images and all
sorts of applications requiring high-capacity array and high speed operation of
Fault-tolerant disk system is implemented on 1 RAID technology, which is a
mirror image discs: data recorded on two or more disks simultaneously. In doing
so, there is more than one copy of the data, and the level of redundancy for
recording information on the two drives already is 100%. With the crash or when
any one of these discs are read with his mirror. The disadvantage of RAID 1 is
the high redundancy of equipment and possible problems with the replacement of a
damaged disk, if implemented RAID 1 functionality programmatically, rather than
through a special controller.
In 2 RAID technology, commercial realization of which are virtually non-existent,
provides for protection of data by correcting codes Hemminga. Records data
distributed on several drives, and control levels ECC (Error-Correction Code)
recorded on one or more designed specifically for this disc. The disadvantage of
RAID 2 is a large share of ECC-disks in the array, which makes implementation of
this architecture is quite expensive. In addition, the controllers of modern
disk devices often have embedded error correction schemes.
In RAID 3 data distributed on disks and information for aggregate data segments,
located in the same sectors on different physical disks, is determined checksum
code or parity, which are recorded on a separate disk. 3 RAID Arrays provide
high speed data transmission in carrying out operations like reading and writing,
and, compared with arrays RAID 2, they need fewer disks for storing parity. The
disadvantage of this level RAID 3 is a sufficient complexity and the feasibility
of hardware only way.
Array RAID 4 increases the productivity of small amounts of data transmission
through parallelism, giving the opportunity to perform more than one treatment
entry to the group concluded in time unit. Logical blocks of information in a
RAID 4 is not distributed among individual discs, and each individual unit lands
on a separate disk. This makes it possible to perform several different requests
for reading simultaneously.
In RAID 5 arrays, as in RAID 4, disks alternately placed large blocks of data,
but, unlike the previous system-level control information among the disk array.
This small change has had a tremendous impact on the productivity of small
arrays of information recording. If the transaction records could be planned so
that to apply for relevant data and blocks them parity to different disks, the
possibility of parallel implementation N / 2 records, where N - the number of
disks in the group. The organization is equally high capacity, recording and
reading as small or large volumes of information, making it the most attractive
in cases of mixed uses.
The following RAID levels designed to improve the reliability of data storage,
but their implementation costs relatively expensive.
RAID 6 is an expanded version of RAID 5, which provides for dual control parity
stored information. If RAID 5 are invited to only one dimension of the matrix
disk, the second dimension of which are sectors, the 6 RAID drives are combined
in a two-dimensional array in such a way that the sector is the third dimension.
Monitoring is carried out as the parity of lines, as in the systems level 5, and
on columns, which in turn can to delaminate to enable parallel recording. If
such an organization can overcome any waivers two disks and many refusals three
disks. However, in carrying out the logical record six hits really happening to
the disk - for the old data, for parity of lines and columns, as well as to
record new data and new values of parity. For some applications with very high
reliability requirements for such a redundancy may be acceptable, but RAID 6
architecture has a very low productivity records in connection with the need to
calculate the amounts and additional controls for traditional supercomputing and
processing transactions, the technique is not suitable.
RAID Level 7 is a corporate decision of Storage Computer. It differs from other
RAID levels that imply an asynchronous components of their work and independence
of management. For this reason, controller RAID disk arrays 7 should run special
real-time operating system. RAID 7 systems have a higher productivity compared
to other levels of RAID arrays. They shared data on disks information, and
reference information is stored on a separate disk. In order to improve
productivity and record the reading centrally cached.
In architectures RAID levels 10 and 0 +1 mirror image drives, as RAID 1 in the
system, coupled with alternate accommodation segments of the data provided in an
array of RAID level 0. These architecture developed when needed more
productivity than that which provides an array of RAID 1. To implement system-level
RAID 0 +1 or 10 requires a minimum of four disk. These systems cost quite
expensive and poorly scaled.
In RAID level 53 architecture combines RAID levels 0 and 3. For its
implementation requires a minimum of 5 discs. RAID 53 alternately writes small
segments of data for the first two drives, and information about parity on the
third disc. The last two drives - the fourth and fifth contain the same data,
alternately recorded large blocks without parity, as is done in the system RAID
RAID technology is constantly evolving and improving. The most expensive system
avtokonfigurirovaniya and acquire the ability to automatically select levels in
real time. They use real-time operating systems, which analyze the flow of data
and determine the most suitable environment for their accommodation. Budget same
system usually focus on a narrow range of tasks and are based on a fixed
standard levels or on the basis of policy decisions. There is also the
introduction of functions RAID on server motherboards, or a combination of
partial implementation at affordable motherboard controller and the motherboard.
The most interesting modern technology implementations build fault-tolerant disk
arrays are Dynamic Migration and AutoRAID.
Dynamic Migration - dynamic migration, is a technology that provides automatic
distribution of data in the system with multiple levels of RAID. Because most
systems for a large part of the data has a low intensity use, the array is
divided into two subsystems - RAID 5 and RAID 1. The data that are used most
heavily, posted on an array architecture RAID 1, while all others at RAID 5.
Thus, the system continually improves its performance by analyzing the activity
data, and that is sufficiently high rate of use of disk space.
AutoRAID technology developed by Hewlett Packard for storage systems and data
processing. Systems based on AutoRAID constitute a fail-safe disk arrays, which
are automatically selected for coding standard RAID data to the maximum extent
appropriate current user queries.
AutoRAID controller, unlike traditional systems, to display the addresses of
blocks of data in the computer addresses of blocks of disk array uses algorithms
dynamic rather than static. Dynamic algorithms allow data to locate any physical
disk array. This makes it possible to broadcast RAID levels from one to another,
with such an operation remains transparent to the host machine. Analyzing the
type and intensity of requests, AutoRAID chooses the most suitable for serving
algorithms, ensuring the highest possible speed possible while maintaining
reliability storage subsystem.
An important feature AutoRAID technology is the possibility of new and
additional disks to balance the system. Under-balanced in this case refers to
the distribution of data on all disks system to achieve maximum performance and
utilization of the disks for active hot reserve.