Upload
ravi-narayan-bhat
View
223
Download
2
Embed Size (px)
Citation preview
8/9/2019 Raid Technology & Troubleshooting
1/49
Confidential
INTRODUCTION TO RAID
The redundant array of Independent Disks (RAID) architectureintegrates multiple disk drives into an array to provide higher performance, capacity and reliability as compared to a single largedrive.
This architecture is also known as Redundant array of InexpensiveDisks and comprises various implementations called RAID levels.
8/9/2019 Raid Technology & Troubleshooting
2/49
Confidential
RAID levels
RAID 0
RAID 1
RAID 2
RAID 3
RAID 4
RAID 5
8/9/2019 Raid Technology & Troubleshooting
3/49
Confidential
RAID-0
RAID 0 represents a striped disk array that does not store parityinformation. As a result, it does not provide data redundancy.
RAID 0 represents a striped array that does not store parity information. Asa result, it does not provide data redundancy.
RAID 0 arrays with large stripes are beneficial for multi-user environments
and RAID 0 arrays with small stripes are used in single-user systems thatrequire access to long and sequential records.
8/9/2019 Raid Technology & Troubleshooting
4/49
8/9/2019 Raid Technology & Troubleshooting
5/49
Confidential
Advantages and Disadvantages of RAID-0
Advantages-
High performance
Handles data redundancy
Disadvantages
Lack of data redundancy
Loss of even one disk in the array result in the loss of the entire volume.
8/9/2019 Raid Technology & Troubleshooting
6/49
Confidential
RAID-1
RAID 1 provides data redundancy by implementing disk mirroring. This isdone by using a pair or group of identical primary and secondary hard disks.
In disk mirroring, data is written concurrently to both the primary andsecondary hard disks.
If the primary hard disk fails the secondary or mirrored hard disk is used
until the primary hard disk is restored. This restoration takes place by usingthe data on the mirrored disk.
8/9/2019 Raid Technology & Troubleshooting
7/49
Confidential
Disk Mirroring
Disk Mirroring offers complete redundancy of data but it is expensive. This
is because twice the disk capacity is required to store the same volume of
data.
Disk mirroring offers better read performance as data can be read from the
hard disks concurrently.
Write operations are slow because the same data is written to both the hard
disks.
8/9/2019 Raid Technology & Troubleshooting
8/49
Confidential
Disk Duplexing
The fault tolerance of disk mirroring can be enhanced by implementing disk
duplexing. Disk mirroring consists of a single disk controller for the mirrored
hard disks in the array.
This makes the data irrecoverable if the controller fails. Disk duplexing
overcomes this problem by utilizing a separate disk controller for the
mirrored disks.
Disk duplexing results in increased costs.
8/9/2019 Raid Technology & Troubleshooting
9/49
Confidential
RAID-3
RAID 3 requires at least three hard disks in which two are striped disks. Thethird hard disk is used to store parity data.
If one striped disk fails, the data can be restored from the from the paritydata.
The write operations in RAID 3 are slow as parity data has to be stored on
the parity disk. The read performance is high because the parity disk is used only when the
data cannot be read from the striped drives.
8/9/2019 Raid Technology & Troubleshooting
10/49
Confidential
Use of RAID-3
The capacity of a RAID 3 array can be increased by adding more hard
disks. This involves an effective increase in the size of the parity disk, which
should be equal to or greater than the size of the individual striped disks.
RAID 3 is most suited for applications that access data sequentially.
It should not be used for intensive and transactional database applications.
8/9/2019 Raid Technology & Troubleshooting
11/49
Confidential
RAID 5
RAID 5 requires at least three hard disks. The parity data is distributedacross all disk drives. As a result, a portion of the total disk space isdedicated for storing parity data.
This portion generally amounts to the size of one hard disk in the array.
RAID 5 enhances read performance as data can be read from all the disk
concurrently. Write operations are slow because parity data is recorded in the parity
stripes.
8/9/2019 Raid Technology & Troubleshooting
12/49
Confidential
Benefits of RAID-5
When a hard disk fails in a RAID 5 array, it can be restored from the parity
information stored on the other hard disks.
Read write operations can continue on these hard disks while the data is
being restored from the information.
This lowers the performance but greatly increases the availability of data in
the array.
RAID 5 very useful for storing critical data.
8/9/2019 Raid Technology & Troubleshooting
13/49
Confidential
RAID 2 and RAID 4
RAID 2 is rarely implemented because it is designed for disk drives
that do not have built-in Error-Correction Code. At present, ECC is
embedded in most disk drives.
RAID 4 is seldom used because it does not support multiple
concurrent write operations.
8/9/2019 Raid Technology & Troubleshooting
14/49
Confidential
8/9/2019 Raid Technology & Troubleshooting
15/49
Confidential
Hybrid RAID level
RAID integrates multiple disk drives into an array to provide
optimum performance, capacity and reliability.
By combining different RAID implementations, various hybrid RAID
levels can be created.
8/9/2019 Raid Technology & Troubleshooting
16/49
Confidential
Hybrid RAID level types
1+5
5+1
0+1
1+0
0+5
5+0
8/9/2019 Raid Technology & Troubleshooting
17/49
Confidential
1+5 and 5+1
It is a combination of disk mirroring and block striping with
distributed parity.
A RAID 1+5 array is created when a number of mirrored sets are
striped with parity.
On the other hand, a RAID 5+1 array is a mirrored configuration of
complete RAID-5 arrays.
8/9/2019 Raid Technology & Troubleshooting
18/49
Confidential
0+1 and 1+0
The 0+1 and 1+0 hybrid levels use mirrored sets in conjunction with thestriped sets without parity.
RAID 0+1 is created by mirroring two striped sets, where as a RAID 1+0array is created by striping multiple mirrored pairs.
RAID 0+1 and 1+0 combine the advantages of striping and mirroring to
obtain large arrays that can provide high performance and efficient faulttolerance.
8/9/2019 Raid Technology & Troubleshooting
19/49
Confidential
8/9/2019 Raid Technology & Troubleshooting
20/49
8/9/2019 Raid Technology & Troubleshooting
21/49
Confidential
8/9/2019 Raid Technology & Troubleshooting
22/49
Confidential
ZCR Cards
RAID levels are implemented using Zero Channel RAID (ZCR) cards. A
ZCR card is a compact implementation of a RAID controller card and
functions with an onboard SCSI chip. This allows for more space on the
card. The compact design of a ZCR card reduces the cost of RAID controller
card. ZCR provides the flexibility to add a card at lower cost that one-
channel and two channel RAID cards.
8/9/2019 Raid Technology & Troubleshooting
23/49
Confidential
8/9/2019 Raid Technology & Troubleshooting
24/49
Confidential
RADIOS
To use a ZCR card, you need a RAID I/O system (RAIDOS) chip embedded
on the motherboard.
This chip is used to enable an embedded I/O controller. The I/O controller
and I/O processor are the main hardware components of a RAID system.
The RAIDIOS circuit allows the I/O processor to configure the controller and
service its interrupts.
8/9/2019 Raid Technology & Troubleshooting
25/49
Confidential
8/9/2019 Raid Technology & Troubleshooting
26/49
Confidential
8/9/2019 Raid Technology & Troubleshooting
27/49
Confidential
Software RAID Vs Hardware RAID
In software RAID, the existing CPU cycles are used. This reduces theprocessing capacity of a server.
In hardware RAID, dedicated circuitry and embedded software are used.
This helps to provide RAID functionality without increasing the load on theCPU.
In both hardware and software RAID, software is used to calculate the arrayallotments and data flow.
8/9/2019 Raid Technology & Troubleshooting
28/49
Confidential
8/9/2019 Raid Technology & Troubleshooting
29/49
Confidential
IOC RAID
The common types of hardware RAID implementations are I/O Controller(IOC) and I/O Processor (IOP). In IOC RAID, processors on disk controllerare used to handle RAID functionality.
IOC based RAID systems can run processes in the background and areindependent of operating systems running on servers.
The IOC performance is constrained by the controllers processor speedand memory bandwidth.
8/9/2019 Raid Technology & Troubleshooting
30/49
Confidential
8/9/2019 Raid Technology & Troubleshooting
31/49
Confidential
IOP RAID
In IOP, various RAID processes are performed by using a special processor
or subsystem in the array. IOP is a commonly implemented hardware RAID
option and is used in servers and Host Bus Adaptors (HBA).
IOP RAID systems have a dedicated processors and firmware that helps inefficient handling of RAID tasks, such as RAID level migration. However,
IOP is an expensive RAID implementation.
8/9/2019 Raid Technology & Troubleshooting
32/49
Confidential
8/9/2019 Raid Technology & Troubleshooting
33/49
Confidential
IOC/IOP RAID Comparison
When compared to IOP based RAID systems, IOC based RAIDsystems are cheaper. However, they lack in performance andflexibility.
For example, IOC does not provide optimum performance whenused with applications such as databases.
Moreover, IOP handles more RAID-management jobs as comparedto IOC.
8/9/2019 Raid Technology & Troubleshooting
34/49
Confidential
8/9/2019 Raid Technology & Troubleshooting
35/49
Confidential
Software RAID: Types
Software RAID implementations are of two types:
RAID based on drivers
RAID based on OS
8/9/2019 Raid Technology & Troubleshooting
36/49
Confidential
RAID-System based on drivers
A driver based RAID system is implemented in disk controller
drivers. Driver based RAID is independent of the operating system
but is dependent on the version of the hard disk controller for RAID
functionality.
8/9/2019 Raid Technology & Troubleshooting
37/49
Confidential
RAID-System based on OS
A RAID system that is based on the OS is implemented by including a RAID
engine in the OS.
The RAID system is independent of the hard disk controller but is
dependent on the OS. For proper functioning, both types of Software RAID are dependent on the
systems processor and memory resources.
8/9/2019 Raid Technology & Troubleshooting
38/49
Confidential
Software RAID: Issues
Software-based RAID is cheaper that hardware based RAID.
However, software RAID systems have certain disadvantages when
compared to hardware RAID systems.
8/9/2019 Raid Technology & Troubleshooting
39/49
Confidential
Software RAID: Issues
Disadvantages include limited performance as less reliability. SoftwareRAID has components that depend on the OS. This results portability issue.
Software RAID solutions are implemented as kernel mode components.Moreover, in OS such as Linux, the RAID solutions are included in Kernel.Usually, kernel mode components require a scheduler to preempt their
operations when their time quantum expires or a higher priority task isscheduled.
8/9/2019 Raid Technology & Troubleshooting
40/49
Confidential
Software RAID: Issues
A kernel mode RAID engine needs to share processor time with other
kernel mode components and applications.
This may lead to an increase in the CPU load, thereby affecting the engines
performance. An increase in the CPU load may be caused due to factorssuch as network traffic, application servers, operating system architecture
and components, and increased I/O processing to and from the secondary
storage.
8/9/2019 Raid Technology & Troubleshooting
41/49
8/9/2019 Raid Technology & Troubleshooting
42/49
Confidential
Hardware RAID advantages
In contrast, hardware RAID offers various advantages. In hardware RAID,
the associated software or firmware is executed on a dedicated processor.
Therefore, hardware RAID does not share the processor with other kernel
mode components and applications.
This provides the advantages of asymmetric multiprocessing such as
distributed task processing.
8/9/2019 Raid Technology & Troubleshooting
43/49
Confidential
Hardware RAID - Advantages
Hardware RAID is not dependent on the operating system of a
server. Moreover, in the case of a RAID hardware malfunction the
server can continue to operate and alert the user about malfunction.
If the server fails, hardware RAID continues to function with the help
of backup modules.
8/9/2019 Raid Technology & Troubleshooting
44/49
Confidential
RAID DISK concepts
Mirrored Drives
Hot Plug Drives
Hot Swap
Hot Spare Drives
Hot Plug Boards
8/9/2019 Raid Technology & Troubleshooting
45/49
Confidential
Mirrored Drives
Most RAID arrays use mirrored drives where all the data on a serveris written simultaneously to two hard disks.
The RAID Controller uses the mirrored drive when one hard disk in aRAID array fails and this ensures that the downtime is kept at aminimum.
The process of a mirrored drive replacing a failed drive is known asfailover.
8/9/2019 Raid Technology & Troubleshooting
46/49
Confidential
Hot Plug Drives
Hot plug drives ensure that a server is always online. A failed drive
in a RAID array can be replaced with hot plug drives.
A hot plug drive can be installed or removed from a RAID array
without interrupting server operations.
8/9/2019 Raid Technology & Troubleshooting
47/49
Confidential
Hot Swap
Hot swap is the ability of a RAID system to recognize a drive that is
inserted while the system is running.
For example, if a drive in a RAID 1 array fails, the malfunctioning
drive can be replaced with a new drive while the system is still
operating.
8/9/2019 Raid Technology & Troubleshooting
48/49
Confidential
Hot Spare Drives
A hot spare drive is connected to a RAID Controller even when other
drives in the RAID array are not functional. When a drive fails, the
controller automatically rebuilds data on the hot spare drive and
disconnects the failed drive.
8/9/2019 Raid Technology & Troubleshooting
49/49
Confidential
Hot Plug Boards
Hot plug boards are used for hard disk failures in a RAID array. A
hot plug board allows compatible hard disk drives and peripherals to
be installed or removed without shutting down the server.
The advantages offered by hot plug board are scalability, high
uptime and minimal down time.