1. Physical Storage Architecture
Introduction
This lesson discusses a number of elements in physical storage architecture. It covers "Just a Bunch of Disks" (JBOD), the importance of Redundant Array of Independent Disks (RAID), the various levels and combinations of RAID, multipathing, and caching.
Importance
Understanding the basics of storage architecture is very important given Cisco's continued movement into the Storage Area Network (SAN) market. The Systems Engineer (SE) needs to be aware not just of Cisco's relevant products, but of how they fit into storage architecture. This lesson goes beyond the physical components of a storage system to how these physical components function as part of a larger whole.
© 2003, Cisco Systems, Inc. All rights reserved. 1

2. Lesson Objective
Upon completion of this lesson, you will explain storage architectures, including JBOD, RAID, multipathing, and caching.
Performance Objective
Upon completion of this lesson, you will explain storage architectures, including JBOD, RAID, multipathing, and caching.
Enabling Objectives
Explain the function of Array Controllers
Explain JBOD
Explain the meaning of the acronym "RAID" and indicate the function of RAID in a storage environment.
Explain where RAID is performed in a system's architecture
Identify the most commonly used RAID levels
Explain RAID level 0 (Data Striping) including its advantages and disadvantages
Explain RAID level 1 (Disk Mirroring), including its advantages and disadvantages
Explain RAID level 5 (striping with distributed parity), including advantages and disadvantages
Explain the RAID level combinations 1+0, 0+1, and 1+5 (Mirrored Array of Striped Disks), including advantages and disadvantages
Explain multipathing, including active/active, active/passive configurations and an iSCSI multipathing implementation
Explain the various roles of multipathing in a storage environment
List multipath products from EMC, Veritas, and HP
Continued…

3. Outline JBOD (Just a Bunch Of Disks) RAID—Adding Redundancy to Storage
Where is RAID performed?
The Most Common Levels of RAID Data Protection
RAID level 0: Data Striping
RAID level 1: Disk Mirroring
RAID Parity
RAID level 5:Striping with distributed parity
Combining RAID levels
Multipath I/O design
Multipathing solution
Multipath products
What is a cache?
Caching implementations.
Cache hits and misses
Common cache algorithms
Summary
Prerequisites
The qualifications to attend the program.

4. JBOD—Just a Bunch of Disks
TRGT 1
TRGT 3
SCSI CABLE
Server
With SCSI
TRGT 0
TRGT 2
TRGT 4
SCSI - JBOD
TRGT 5
TRGT 4
TRGT 3
JBOD–Just a Bunch of Disks
Objective
Explain JBOD.
Introduction
This section defines and introduces JBODs.
Facts
JBOD is the acronym for "Just a Bunch Of Disks." A JBOD consists of two or more disks packaged in the same enclosure and connected to a system or server. Originally used to mean a collection of disks that is coordinated by control software, the term JBOD now refers to a cabinet of disks where RAID functionality is not present. The internal connections typically use either the Small Computer Systems Interface (SCSI) bus or Fibre Channel Arbitrated Loop (FCAL) (see the above diagram). From a system's perspective, it sees only the disk drives, although the enclosure may provide monitoring or management facilities.
Higher end JBODs include support for hot swap, as well as redundancy of power supplies, fans, controllers and dual ported disks.
Practice Items
Is there any advantage to a JBOD not configured for RAID over the same disks configured for RAID?
FC-AL
Fiber CABLE
Server
With FC-AL
TRGT 2
TRGT 0
TRGT 1
FC–AL—JBOD

5. RAID—Adding Redundancy to Storage
RAID—Redundant Array of Independent Disks
RAID is a family of techniques for managing multiple disks to deliver desirable cost, data availability, and performance characteristics to host environments.
If a device fails, RAID ensures that the data can be retrieved or reconstructed.
RAID—Adding Redundancy to Storage
Objective
Explain the meaning of the acronym "RAID" and indicate the function of RAID in a storage environment.
Introduction
This section explains the acronym RAID and introduces its importance in a storage environment.
Definition
A RAID combines a group of independent disks into an array of disks that provides performance levels that exceed those of a single large drive. The array is viewed as a single logical device. The performance characteristics of the collection of disks is determined by the specific architecture of the array. There are several architectures for disk arrays. These are referred to as the RAID levels.
Facts
RAID is the acronym for “Redundant Array of Independent Disks”. The "I" is sometimes said to stand for "Inexpensive“.
RAID is a family of techniques that utilize duplication and/or specialized placement of data across multiple disks. Depending upon the techniques used, performance and/or fault tolerance can be enhanced.
RAID can allow a system to continue operating, without loss of data, even after a hard disk crash.

6. RAID—Adding Redundancy to Storage (cont.)
Server
JBOD
I/O Interface
Disk Failure!
All data on failed disk lost
Without RAID
RAID—Adding Redundancy to Storage (cont.)
Facts
RAID addresses two performance issues: the speed with which data can be written to storage, and the ability to continue reading and writing data when a disk and/or disk subsystem fails (redundancy). For situations requiring speed, RAID solutions provide the ability to write data to multiple disks simultaneously. In situations where a data access is critical, RAID solutions can provide a number of redundancy strategies. Each of these strategies allows one or more disk drives to fail without losing the ability to access data or write data. The redundancy is based on the technique of writing redundant information on other disks so that if a disk fails the data associated with the failed disk can be recovered using the redundant information on other disks.
The RAID functions may be performed by software in the host system (software RAID or host-based RAID, or by a RAID controller (which may be internal to the system or server or external).
Example
At any time of the day a person can pull up to a gas station, insert a credit card into a card reader to pay for the gas, fill up a car, and be on their way. How can the credit card company ensure that their network and storage are available every hour of every day? RAID solutions can provide the needed storage database redundancy.
Continued …

7. Where is RAID Performed?
The RAID function may be performed at one of several places.
Disk Subsystem
Host
Host Interface
Power Supply
Software
Controller
Cache
Switch
Where RAID is Performed
Objective
Explain where RAID is performed in a system's architecture.
Introduction
This section introduces the locations within a system that the RAID function can be performed.
Facts
The RAID function may be performed at a number of different places:
It may be performed within a storage subsystem—a disk array or RAID array with one or more RAID controllers.
By an add-in card in the host system or server.
By software running in the host system or switch. Some operating systems provide built-in support for RAID functionality.
The RAID function (wherever it is located) is connected to a bunch of disks. The disks may be integrated within the same enclosure as the RAID function or externally (JBOD).
Internal I/O Channel
HBA

8. The Most Common Levels of RAID Data Protection
0 Data striping
1 Disk mirroring
5 Striping with distributed parity Mirrored Array of Striped Disks
0 + 1 Striped Array of Mirrored Disks
1 + 5 Striped Array of Mirrored Disks
The Most Common Levels of RAID Data Protection
Objective
Identify the most commonly used RAID levels.
Introduction
This section introduces the most commonly implemented RAID levels.
Facts
There are several ways in which disks may be combined into arrays and redundancy provided. These are referred to as the RAID levels.
The list included is not complete, but it does reflect the most commonly implemented RAID levels:
0—Data Striping
1—Mirroring
5—Data Striping with Parity
1+0—Mirrored Stripes
0+1—Striped Mirrors
1+5

9. RAID Level 0: Data Striping
"RAID" 0 is really "AID", as there is no redundancy.
Data Stripe A0 A1 A2 A3 B0 B1 B2 B3 C0 C1 C2 C3 D0 D1 D2 D3
RAID Level 0: Data Striping
Objective
Explain RAID level 0 (Data Striping) including its advantages and disadvantages.
Introduction
This section discusses the basic features of RAID 0.
Facts
RAID level 0, or Disk Striping, is the process of breaking data into blocks of predetermined size and writing the blocks to separate disk drives.
Example
The diagram displays four physical drives that have been combined into an array. Each drive is sufficient in size to accommodate four stripes. If the individual blocks were 512K in size then a single stripe file of 4 X 512K could be written across a stripe in one Input/Output (I/O) cycle. By accessing the drives in parallel, the effective transfer rate is equal to the sum of the transfer rates of all of the drives in the RAID set.
There is a major risk associated with simple data striping, if a disk drive fails all of the data stored on the array is lost. Giving up redundancy increases performance, because there is no parity calculation. This is important for those seeking the best possible performance.
Continued….
Data is distributed across the disks. Each disk contains only a portion
of the data. RAID level 0 does not provide any redundancy. If a disk fails,
all data is lost.

10. RAID Level 1: Disk MirroringB0 B1 C0 C1 D0 D1 = A1 F1 G1 H1 E1 E0 F0 G0 H0
RAID Level 1: Disk Mirroring
Objective
Explain RAID level 1 (Disk Mirroring), including its advantages and disadvantages.
Introduction
This section discusses the basic features of RAID 1.
Facts
Disk Mirroring (RAID level 1) maintains identical copies of the same data on two separate disk drives. When a write occurs, the data is written to both drives. Duplication of data on the second drive can be done using either a hardware RAID controller or software.
The characteristics of data I/O do not change much with disk mirroring. When data is being written, it must be written to both drives. When a read occurs, data may be read from either drive. The ability to read the data from either drive provides the redundancy in disk mirroring. If one of the drives fail, the data is still available from the other drive. It can also employ duplexing, in which the controller card is duplicated as well as the drive, thus providing tolerance against failures of either a drive or a controller.
Continued…
Data is duplicated on a second disk. If one disk of the mirrored set fails,
data can be read from the other disk.

11. Providing redundancy via parity
RAID Parity
Providing redundancy via parity
Parity is computed by an exclusive-or (XOR) operation.
If A XOR B XOR C = Parity, then A XOR B XOR Parity = C
RAID Parity
Objective
Explain RAID parity.
Introduction
This section introduces RAID Parity, its advantages and disadvantages, and contexts in which it might be used.
Facts
Mirroring provides protection against data loss due to a device failure, but with significant overhead as two copies of the data are recorded—this equals 100 percent overhead.
Parity is a parity technique used to provide redundancy with less overhead. Parity is computed by an exclusive-or (XOR) operation. If n data bytes are XOR'd to create a parity byte, then any one of the data bytes can be recreated by XOR'ing the remaining bytes with the parity byte.
If A XOR B XOR C = Parity, then A XOR B XOR Parity = C. This property is used by RAID levels 3, 4, 5, and 6.
Optional Practice
Given the following 3 bytes, calculate the value of the parity byte:
A –
B –
C –
P - _ _ _ _ _ _ _ _ 1 1

12. RAID 5: Striping with Distributed Parity
Data
with Parity
Data
with Parity A0 B0 C0
Parity 0 A0 B0 C0
Parity 0
Parity 1
Parity 1 A1 A1 B1 B1 C1 C1 C2 C2
Parity 2
Parity 2 A2 A2 B2 B2 B3 B3 C3 C3
Parity 3
Parity 3 A3 A3
RAID Level 5: Striping with Distributed Parity
Objective
Explain RAID level 5 (striping with distributed parity), including advantages and disadvantages.
Introduction
This section discusses the basic features of RAID 5.
Facts
RAID 5 uses a combination of block-level striping and distributed parity information. RAID 5 writes the data to one or more of the data disks in the RAID set, and writes the parity information to the other disk drive in the set. For example, in the illustration, a file (represented by blocks A0, BO, and C0) is written across the first three drives of the stripe, and its parity information is written to the fourth drive. Data protection is provided through the use of the parity block.
While every data write also causes a write to the parity disk, congestion is not a problem. This is due to the fact that the parity information is distributed among the disks. How does the parity help reconstruct the data in the event of a disk failure? Consider the example again. The data for the first stripe is written to the disk. Imagine that the parity value is calculated as A0 XOR B0 XOR C0 = Parity. If the second disk fails then the data on that disk that is associated with the first stripe could be reconstructed by using the calculation of A0 XOR C0 XOR Parity0 = B0. This will result in a loss in read performance, but the data is still available to applications.
As disk prices have come down, and capacities have increased, RAID 5 has become less popular.
Continued…
Data is written on a data disk. Parity is computed and written on one
of the data disks. If one of the data disks fails, the data can be
reconstructed by XORing data from the other data disks with the parity.

13. RAID 5: Striping with Distributed Parity (cont.)
RAID write penalty:
When performing a write command, a series of I/O operations must take place
The RAID controller must perform four I/O operations per write update resulting in a significant performance hit
RAID Level 5: Striping with Distributed Parity (cont.)
Facts
Note the I/O multiplication associated with RAID 5 (and 4 and 6).
RAID levels 4, 5, and 6 incur additional overhead on write updates due to the need to update the parity data. When a write occurs, the RAID controller must perform a series of I/O operations to write the new data and update the parity data.
When performing a write command a series of I/O operations must take place. The RAID controller:
Reads the old data from the data disk
Writes the new data to the data disk
XOR’s the old data with the new data to compute an intermediate parity
Reads the old parity from the parity disk
XOR’s the old parity with the intermediate parity computed above
Writes the new parity to the parity disk
This results in four I/O operations per write update.

14. Combining RAID Levels RAID 1+0: Mirrored Array of Striped Disks
Mirrored Set – RAID Level 1
Striped Set
RAID level 0
Striped Set
RAID level 0 = A0 A1 A2 A0 A1 A2 = B0 B1 B2 B0 B1 B2 = C0 C1 C2 C0 C1 C2
Combining RAID Levels
Objective
Explain the RAID level combinations 1+0, 0+1, and 1+5 (Mirrored Array of Striped Disks), including advantages and disadvantages.
Introduction
This section discusses the basic features of RAID 1+0.
Facts
RAID 1+0 is a combination of RAID 0 and RAID 1. The configuration contains a stripe that is mirrored to a second stripe. This configuration allows striping to be used to provide high I/O performance, and mirroring to be used for fault tolerance.
Advantages
The advantages of RAID 1+0 include:
Provides fault tolerance
High I/O rates
Disadvantages
The disadvantages of RAID 1+0 include:
The need to have two sets of drives makes the cost of implementation high
Limited scalability = D0 D1 D2 D0 D1 D2
The highest-level RAID function is level 1 (mirroring). This provides
redundancy in case of a disk failure. The lower level uses RAID level 0
(striping) to provide high performance.
RAID 1+0: Mirrored Array of Striped Disks

15. Combining RAID Levels (cont.)
Mirrored Set
RAID level 1
Striped Set
RAID level 0 = A0 A0 A0 A1 A1 = A1 A2 A3 A2 = A2 A4 A5 A3 = A3 A6 A7
Combining RAID Levels (cont.)
Facts
RAID 0+1 combines RAID 1 and RAID 0. The highest-level RAID function is level 0 (striping). This provides the high performance associated with RAID level 0. Instead of striping the data across individual disk, the lower level uses RAID level 1 (mirroring) to provide redundancy and protect against data loss if a disk fails. This results in the data being simultaneously mirrored and striped. This configuration is used to provide high reliability (mirroring) combined with high I/O performance (striping).
Advantages
The advantages of RAID 0+1 include:
Provides fault tolerance
High I/O rates
Disadvantages
The disadvantages of RAID 0+1 include:
Very expensive
High overhead
Limited scalability
Applications that can benefit from RAID 0+1
RAID 0+1 can be beneficial to applications that require both high performance and reliability. The types of applications that may benefit from RAID 0+1 include:
Database servers requiring high performance and fault tolerance A3 A3 A6 A7
RAID 0+1: Striped Array of Mirrored Disks

16. Combining RAID Levels (cont.)
Mirrored Set – RAID Level 1
Data
with Parity
Data
with Parity A0 B0 C0
Parity 0 A0 B0 C0
Parity 0
Parity 1 A1 B1 C1
Parity 1 A1 B1 C1 C2
Parity 2 A2 B2 C2
Parity 2 A2 B2
Combining RAID Levels (cont.)
Facts
RAID 1+5 combines RAID 5 and RAID 1. This results in a data stripe set with parity being mirrored to a second data stripe set with parity. This configuration is used to provide extremely high fault tolerance combined with good performance. The performance is good but not very high for the cost involved, nor relative to that of other multiple RAID levels. The fault tolerance is the primary benefit of this strategy; an eight-drive RAID 51 array can tolerate the failure of any three drives simultaneously, and even as many as five, as long as at least one of the mirrored RAID 5 sets has no more than one failure.
Advantages
The advantages of RAID 1+5 include:
Provides fault tolerance
Good I/O rates
Disadvantages
The disadvantages of RAID 0+1 include:
Very expensive
High overhead
Complex implementation
Limited scalability
Applications that can benefit from RAID 1+5
RAID 1+5 can be beneficial to critical applications requiring very high fault tolerance. B3 C3
Parity 3 A3 B3 C3
Parity 3 A3
RAID 1+5: Striped Array of Mirrored Disks

17. Multipath I/O Design A redundant I/O design
Dynamic failover and recovery
Host with dual FC HBAs and Multipathing Software Installed
HBA Driver
Multipathing
Application
Storage Array with Redundant Controller Ports
Multipath I/O Design
Objective
Explain multipathing, including active/active, active/passive configurations, and an iSCSI multipathing implementation.
Introduction
This section introduces multipathing, including a definition and an overview of multipathing solutions.
Facts
Multipathing is a form of communication path control that provides dynamic multiple hardware paths to a storage array (LUN). The current generation of multipath storage solutions is supplied by storage subsystem providers, storage software providers, and OS platform providers. Multipath storage products can provide a large spectrum of features and functions that affect the performance, availability, accessibility, configurability, and serviceability of the storage subsystem and system I/O.  
In general, storage hardware providers offer sophisticated multipath storage management solutions that are optimized for their own subsystems and are implemented as server device drivers in order to provide highly granular control over the I/O sent to the storage devices. These solutions are complex and cannot be used with other storage hardware. Examples include IBM Storage Device Driver (SDD), EMC Powerpath, and Hitachi Dynamic Link Manager. 
Subsystem neutral, host-based solutions are often simpler, less-expensive solutions that primarily offer dynamic fail-over and load balancing. An example of a cross-platform, host-based multipathing solution is Veritas Dynamic MultiPathing.
FC Switches

18. Multipath I/O Design (cont.)
Multipathing: Multiple hardware paths to a single drive (LUN)
Active/Active: Balanced I/O over both paths (implementation specific)
Active/Passive: I/O over primary path—switches to standby path upon failure
HBA Driver
Multipathing
Application
Active
Active
Load Balancing
HBA Driver
Multipathing
Application
Primary Path
Multipath I/O Design (cont.)
Facts
Multipathing solutions can be limited to simple failover where a standby path is unused unless or until the primary path fails. This is referred to as an active/passive configuration.
More sophisticated solutions utilize and balance traffic across more than one path as well as providing a failover to the remaining path(s) should one fail.  Such a configuration is referred to as active/active.
Active
Passive
Multipathing Software Monitors Active I/O Path
Standby (Failover) Path
HA Multipathing Solutions

19. Multipath I/O Design (cont.)
iSCSI design (multipathing)
Host with Multiple (iSCSI) NICs and Multipathing Software Installed
iSCSI Driver
Multipathing
Application
Storage Array with Redundant Controller Ports
iSCSI Router ‘A’
Multiple Ethernet Switches
Multipath I/O Design (cont.)
Facts
The same level of high availability fault tolerance design can be accomplished in iSCSI environments. In this scenario, instead of using expensive FC Host Bus Adapters (HBAs) in the server(s) standard 10/100 Ethernet cards are attached to redundant Ethernet switches. The Ethernet LAN is then connected to the FC SAN by way of dual HA iSCSI routers. Multipathing software on the application host provides the same active/active or active/passive monitoring and failover capabilities discussed previously.
Redundant Fabrics
iSCSI Router ‘B’

20. Multi-Pathing Solution
Manages multiple paths to a storage system
Intelligent path selection—Load Balancing
Increased Performance and Throughput
Automatic Error Detection and Path Failover
Application and Business Continuance
Dynamic Recovery
Minimizes System Administrative Planning
Dynamic Configuration
Application, Database, File System, Management Utilities and Volume Manager Independent
Multi-pathing Solution
Objective
Explain the various roles of multipathing in a storage environment.
Introduction
This section introduces the roles of multipathing in a storage environment.
Facts
Multipathing software provides the intelligence and monitoring capabilities needed to implement load balancing, error detection, path failover and path recovery.
In an active/active multipathing configuration, load balancing increases throughput by distributing I/O requests across more than one active I/O channel.
Automatic error detection, path failover and dynamic recovery capabilities help minimize business disruptions caused by loss of data communications.
Sophisticated multipathing implementations ease the burden on system administrators by enabling automatic detection and configuration of new devices. Multipathing is transparent to the applications running on the server.
Practice
Explain the benefits of multipathing from the point of view of IT management.

21. Multipath Products EMC PowerPath Veritas Dynamic MultiPathing
HP StorageWorks Secure Path
and others …
Multipath Products
Objective
List multipath products from EMC, Veritas, and HP.
Introduction
This section introduces multipathing products from EMC, Veritas, and HP.
Facts
The following are a few of the many multipathing solutions available:
EMC PowerPath: This is a server resident software solution that integrates multi-path I/O capabilities, automatic load balancing and path failover functions for servers connected to EMC Symmetrix and CLARiiON storage subsystems. Server support is provided for most UNIX, Windows and Linux operating systems. One server can include up to 32 I/O paths to storage. PowerPath can be included with clustering software such as IBM HACMP, Microsoft Cluster Server and Veritas Cluster Server.
Veritas Dynamic Multipathing (DMP): DMP is included as part of the Veritas Foundation Suite for UNIX, Linux and Windows servers. It supports multiple paths to multiple arrays in both fabric and loop environments. The many-to-many relationship removes potential downtime caused by path, switch, HBA and/or interface card failure. An advantage of DMP is that it works with storage and server hardware from multiple vendors
HP StorageWorks Secure Path: This is a family of high availability multi-pathing software products providing continuous data access from HP's RAID Array to host servers running Windows, UNIX and Novell operating systems. Redundant hardware, advanced RAID technology and Secure Path's automated failover capability are used to enhance fault tolerance and availability. Secure Path effectively eliminates controllers, disk drives, interconnect hardware and host bus adapters as single points of failure in the storage subsystem.

22. Cache is high-speed semiconductor memory that holds portions of data
What is a Cache?
Cache is high-speed semiconductor memory that holds portions of data
Cache is a repository for data retrieved from disk, used to improve performance
What is a Cache?
Objective
Explain caching, including the reason optimum cache performance affects overall system performance
Introduction
This section defines caching and indicates how it improves system performance.
Facts
Caching is a technique used to improve performance of disk subsystems by storing frequently used data in memory. When a request is made to read data, the cache function determines if the requested data is currently contained in the cache. If it is, the data is sent from the cache avoiding the overhead associated with accessing the data from the physical disk.
If the data is not in the cache, the cache function reads the data from the disk. This takes longer than transferring the data from the cache and results in lower performance. Because the cost of memory is significantly greater than the cost of disk storage, the amount of cache memory is limited. Typically, the size of the cache is 1 percent–2 percent of the size of the disk space. Considerable effort goes into developing and tuning algorithms to maximize the number of cache hits.
Why is this important? SEs need to be aware that performance can be related to cache performance, not switches or controllers.

23. Caching Implementations
Host operating system buffer cache
Caching Implementations
Objective
Explain the different ways caching is implemented in storage environments.
Introduction
This section explains basic cached storage implementations.
Facts
Caching can be performed by the operating system in the host or server, or by an array controller in a storage subsystem. While most disk drives provide buffering for data transfers, this memory is not really used as a cache, but rather as a speed matching buffer.
The operating system in a host or server performs caching by allocating memory buffers to hold recently used data. Before attempting to read data from a storage device, the local buffer cache will be checked.
Storage subsystems also perform caching. Typically this is done by storage subsystems with hardware RAID controllers.
Storage subsystem cache

24. Cache Hits and Misses
When a read request occurs, there can be one of two outcomes:
Cache hit:
The requested data is present in the cache and the read can be satisfied directly from the cache without having to access the physical device—measured in microseconds.
Cache miss:
The data is not present so the read can be satisfied only be accessing the physical device—measured in milliseconds.
Cache hit ratio = Percentage of reads satisfied from cache
Cache Hits and Cache Misses
Objective
Explain what a cache hit and what a cache miss is, and the cache hit ratio
Introduction
This section explains what a cache hit or miss is, and the cache hit ratio.
Facts
Cache hit: The requested data is present in the cache and the read can be satisfied directly from the cache without having to access the physical device. When a request is made to read data, the cache function determines if the requested data is currently contained in the cache. If it is, the data is sent from the cache avoiding the overhead associated with accessing the data from the physical disk.
Cache miss: The data is not present so the read can be satisfied only by accessing the physical device. This takes longer than transferring the data from the cache and results in lower performance.
Cache hit ratio: The percentage of the time that read requests can be satisfied from the cache. A 95 percent cache hit ratio means that 95 percent of the reads were satisfied from the cache and only 5 percent had to go to the physical device. A cache hit ratio of 90 percent or higher is certainly considered optimal.

25. Common Cache Algorithms
Read ahead cache
Write back (fast write) cache
Extended adaptive cache
Least recently used
Common Cache Algorithms
Objective
Explain common cache algorithms including read ahead, write back, IBM's extended adaptive cache, and LRU algorithms.
Introduction
This section explains how four common cache algorithms work.
Facts
Read ahead cache: This refers to the strategy of sequentially reading more data from disk than explicitly requested by the application. Applications that tend to read data sequentially can realize significant performance improvements from this strategy.
Write back cache: Also known as “fast write cache”. This improves write performance by allowing the controller to acknowledge completion of a write operation upon depositing data into its cache. Later, the data is committed to disk. High-end implementations use non-volatile RAM or uninterruptible power supply (UPS) mechanisms to insure data integrity.
Extended adaptive cache: Developed by IBM, this is an intelligent cache management strategy that is controlled at the storage subsystem level. It utilizes advanced algorithms in an attempt to predict future disk access patterns
Least Recently Used (LRU): Whenever a cache write threshold is exceeded, an LRU replacement algorithm is employed to de-stage the cache. Once enough free cache exists, cache writes continue.

26. Where RAID is Performed Practice
Label the places where RAID can be performed in a storage system.
Disk Subsystem
Host
Host Interface
Power Supply
Cache
Where RAID is Performed
Instructions
List and explain where RAID can be performed in a storage system.
Internal I/O Channel

27. RAID Practice 1 5 0 + 1 1 + 5 RAID type Description (e.g. mirroring)
Explanation
Advantages/
Disadvantages
Applications 1 5
0 + 1
1 + 5
RAID
Instructions
Complete the matrix. Make your answers brief and to the point. You may need to use a separate sheet of paper to record your answers.

28. RAID Practice (cont.)
In each of the following examples, what is the maximum number of drives that can fail without loss of data?
RAID 0
RAID 5
RAID 1
RAID
Instructions
See above.
RAID 1+5

29. Multipath I/O Design Practice
HBA Driver
Multipathing
Application
Standby (Failover) Path
Primary Path
Active
Passive
Explain multipathing, active/active and active/passive configurations as displayed in the diagrams.
Multipath I/O Design
Instructions
Explain multipathing, active/active and active/passive configurations as displayed in the diagram.

30. Caching Practice
On a sheet of paper, record your answers to the following questions.
What is caching?
Where is caching implemented?
Describe the relationship of the cache hit ratio to system performance.
What are four common cache algorithms?
Caching

31. RAID—Adding Redundancy to Storage Practice
Answer the following questions:
What two performance issues does RAID address?
What is the basic strategy by which RAID achieves redundancy?
RAID–Adding Redundancy to Storage

32. Multipath I/O Design Practice
Answer the following question. Be prepared to share your answers with the class
Explain how iSCSI can help achieve high availability and fault tolerance in a storage environment.
Multipath I/O Design (cont.)

33. Lesson Practice
Define each of the following and give at least two sentences outlining what you feel is important about each topic. Record your answers and be prepared to share them with the class.
JBOD
RAID
Multipathing I/O
Cache
Physical Storage Architecture

34. Summary This lesson presented these key points:
The following are the basic components of physical storage architecture:
JBOD
RAID = Redundant Array of Independent(or Inexpensive) Disks
Common RAID types, including single RAID levels, and combinations of RAID levels.
Caching implementations, cache hits and misses and cache algorithms
Physical Storage Architecture
Review
This lesson discussed a number of elements in physical storage architecture. It covered "Just a Bunch of Disks" (JBOD), the importance of Redundant Array of Independent Disks (RAID), the various levels and combinations of RAID, multipathing, and caching.