RAID

Dr. E. Papanasam
SoC
 Impact of I/O on System Performance
• Suppose we have a benchmark that executes in 100 seconds of
elapsed time, of which 90 seconds is CPU time and the rest is I/O
time
• Suppose the number of processors doubles every two years, but
the processors remain the same speed, and I/O time doesn’t
improve
• How much faster will our program run at the end of six years?
• Elapsed time = CPU time + I/O time; I/O time = 10 seconds
 Impact of I/O on System Performance
• The improvement in CPU performance after six years is 90 /11 =
8
• However, the improvement in elapsed time is only 100/21 = 4.7
• I/O time has increased from 10% to 47% of the elapsed time

• Parallel revolution needs to come to I/O as well as to

computation
• Accelerating I/O performance was the original motivation of disk
arrays
• In the late 1980s, the high performance storage of choice was
large, expensive disks
• By replacing a few large disks with many small disks,
performance would improve because there would be more read
heads
• Good match for multiple processors as well
• Many read/write heads mean the storage system could support
many more independent accesses as well as large transfers
spread across many disk
• Both high I/Os per second and high data transfer rates

• Disk arrays could make reliability much worse

• These smaller, inexpensive drives had lower MTTF ratings than
the large drives
• More importantly, by replacing a single drive with, say, 50 small
drives, the failure rate would go up by at least a factor of 50!
• The solution was to add redundancy so that the system could
cope with disk failures without losing information
• By having many small disks, the cost of extra redundancy to
improve dependability is small, relative to a few large disks
• Dependability was more affordable if you constructed a
redundant array of inexpensive disks
• Led to its name: redundant arrays of inexpensive disks (RAID)
• An organization of disks uses an array of small and
inexpensive disks so as to increase both performance and
reliability

• How much redundancy do you need?

• Do you need extra information to find the faults?
• Does it matter how you organize the data and the extra check
information on these disks?
 RAID
• Rate in improvement in secondary storage performance has
been considerably less than the processors and main memory
• This mismatch has made the disk storage system the main focus
of concern in improving overall computer system performance
• Additional gains in performance by using multiple parallel
components

• Leads to the development of arrays of disks that operate

independently and in parallel
• With multiple disks, separate I/O requests can be handled in
parallel, as long as the data reside on separate disks
• A single I/O request can be executed in parallel if the block of
data to be accessed is distributed across multiple disks
 RAID (2)
• With the use of multiple disks, there is a wide variety of ways in
which the data can be organized and in which redundancy can be
added to improve reliability
• RAID (Redundant Array of Independent Disks) is a standardized
scheme for multiple-disk database design
• Seven levels - zero through six
• Levels
– Do not imply a hierarchical relationship
– Designate different design architectures
 Common characteristics of RAID
• Three common characteristics
– 1. RAID is a set of physical disk drives viewed as a single
logical drive by OS
– 2. Data are distributed across the physical drives of an array
– 3. Redundant disk capacity is used to store parity
information, which guarantees data recoverability in case
of a disk failure
• Details of the second and third characteristics differ for the
different RAID levels
• RAID 0 and RAID 1 do not support the third characteristic
 RAID (3)
• The RAID strategy employs
– multiple disk drives and
– distributes data in such a way enabling simultaneous
access to data from multiple drives
– improving I/O performance
– allowing easier incremental increases in capacity
– Improve access time and improve reliability
• RAID - Redundant Array of Inexpensive Disks
• Inexpensive - small relatively inexpensive disks in the
RAID array alternative to a single large expensive disk
 RAID 0 (Non Redundant)
• Not a true member of the RAID family - It does not include
redundancy
• Data stored in the logical disk in the form of strip
• Striping - Allocation of logically sequential blocks to separate disks
to allow higher performance than a single disk can deliver
• One logical disk is divided into four physical disks
• Round Robin method
• Array management software carry out distribution of data from
logical disk into physical disks
• High I/O request rate - Requests can be handled in parallel (ala
data resides in multiple disks) - Reduces the I/O queuing time
• High data transfer capacity - Request can be executed in parallel (if
the block of data to be accessed is distributed across multiple
disks) - reducing the I/O transfer time
RAID 0
 RAID 1 (Mirroring or Shadowing)

• Traditional scheme for tolerating disk failure

• Uses twice as many disks as does RAID 0
• Data striping as in RAID 0
• Read from either-A read request can be serviced by either of the
two disks
 RAID 1
• Write to both
– A write request requires both corresponding strips be
updated - can be done in parallel
• Recovery from a failure is simple
– If a disk fails, the system just goes to the “mirror” and reads its
contents to get the desired information
• Expensive
– Requires twice the disk space of the logical disk that it
supports
– RAID 1 configuration is limited to drives that store system
software and data and other highly critical files
• Achieve high I/O request rates if the bulk of the requests are
reads
• No significant performance gain over RAID 0 if a substantial
fraction of the I/O requests are write requests
RAID 2 (Redundancy through Hamming code)
• Use of a parallel access technique
• All disks participate in the execution of every I/O request
• Data striping – Bit level striping
• An error-correcting code is calculated across corresponding bits on
each data disk
• Bits of the code are stored in the corresponding bit positions on
multiple parity disks
• Hamming code is used, which is able to correct single-bit errors and
detect double-bit errors
 RAID 2
• Requires fewer disks than RAID 1
• Number of redundant disks is proportional to the log number
of data disk
• If there is a single-bit error, the controller can recognize and
correct the error instantly, so that the read access time is not
slowed
• On a single write, all data disks and parity disks must be
accessed for the write operation
 RAID 3 (Bit-Interleaved Parity)
• Bit level striping
• Parity rather than hamming code
• Simple parity bit for each set of corresponding bits
• Only one redundant disk, no matter how large the array
• Data on failed drive can be reconstructed from surviving data and
parity info
• Read - all disk
• Write - all disk including the redundant parity disk to be updated
• Only one request serviced at a time
 RAID 3
• In the event of a drive failure, the parity drive is accessed and data
is reconstructed from the remaining devices
• Once the failed drive is replaced, the missing data can be restored
on the new drive and operation resumed
• Array of five drives X0 through X3 contain data and X4 is the parity
disk
• The parity for the ith bit is

• Suppose that drive X1 has failed

• If we add to both sides of the preceding equation

• Contents of each strip of data on X1 can be regenerated from the

contents of the corresponding strips on the remaining disks in the
array
 RAID 4 (Block level Parity)

• Block level striping

• Block – one or more word
• Each disk operates independently
• Separate I/O requests can be satisfied in parallel
• Provided read request smaller than striping unit (Block)
 RAID 4
• Write request – Requested data block as well as parity block to
be updated
• More suitable for applications require high I/O request rates
• Small write request require four disk I/O (write penalty)
– One to write a new date
– Two read
• One to read the old data and another to read old parity
to compute new parity
– One to write the new parity
– Each strip write involves two reads and two writes (Read-
Modify-Write procedure)
 RAID 4
• Suppose that a write is performed that only involves a strip on
disk X1.
• Initially, for each bit i, we have the following relationship
Small write update on RAID 3 versus
RAID 4
 RAID 5(Block level Distribute Parity)
• One drawback is parity disk must be updated on every write, so
the parity disk is the bottleneck for back-to-back writes
• Same as RAID 4
• Parity strips are distributed across all disks
• Parity strip allocation – Round Robin scheme
• Avoids potential I/O bottleneck found in RAID 4
 RAID 4 Vs RAID 5
• RAID 5 allows multiple writes to occur simultaneously as long
as the parity blocks are not located on the same disk
 RAID 6 (Dual Redundancy)
• Two different parity calculations
• P - Odd-even parity and Q - Read Solomon code
• Stored in separate blocks on different disks
• No of disks required = N+2 (N - number of disk required for data)
• Possible to regenerate data even if two disks containing user data
fail
• Substantial write penalty, because each write affects two parity
blocks
• hot-swapping
Replacing a hardware component while the system is running
• standby spares
Reserve hardware resources that can immediately take the
place of a failed component
 Summary of RAID Levels
• Data transfer capacity (ability to move data)
• I/O request rate (ability to satisfy I/O requests)