Storage Fundamentals

7/30/2019 Storage Fundamentals

1/99

- 1 -

Storage Basics & EMC Storage


2/99

- 2 -

Day 1

Basics of Storage Technology

Managing and Monitoring the Storage Devices

Business Continuity

Day 2

Switches and Directors

Host Integration with the Storage Devices

Labs

Agenda


3/99

- 3 -

Types of Storage Connectivity

Direct Attached Storage (DAS)

Storage Area Networks (SAN)

Network Attached Storage (NAS)

SAN Foundation

Multipath & Failover

Basics of Storage Technology


4/99

- 4 -

Options for connecting computers (Hosts) to storage.

DAS (Direct Attached Storage): Storage is directly attached by a cable to

the Host.

SAN (Storage Area Network): Storage resides on a dedicated network,

providing any-to-any connection between hosts and storage.

NAS (Network Attached Storage): Storage is attached to a TCP/IP based

network (LAN or WAN), and accessed using CIFS and NFS protocols for

file access and file sharing. A NAS device is sometimes also called a fileserver. It receives request over a network and has an internal processor

which translates that request to the SCSI block I/O commands to access

the appropriate device only visible to the NAS product itself.

Types of Storage Connectivity


5/99

- 5 -

Direct Attach Storage


6/99

- 6 -

Direct Attached Storage is restricted to access by a singlehost.

Sometimes by two or more hosts in cluster (failover orfailback) configurations.

DAS may initially appear to be low cost from the point ofview of each user or department.

However, from the wider perspective of the entireorganization, DAS may be higher for networking approachesdue to the difficulty of sharing unused capacity with otherhosts, and also lack of a central point of management formultiple storage systems

Direct Attach Storage (Cont)


7/99- 7 -

Storage Area Network

SAN serves to interconnect storagerelated resources that are connected tomultiple servers. SAN are usually built

using Fiber Channel technology, butthe concept of a SAN is independent ofthe underlying type of network we canalso use iSCSI (IP based SANs) in

production environment. I/O requeststo disk storage on a SANare called block I/Os because, just asfor direct-attached disk, the read and

write I/O commands identify a specificdevice (disk drive or tape drive) and inthe case of disks, specific block

(sector) locations on the disk.


8/99- 8 -

Access: Longer distance between the processors and storage, higher availability,

improved performance. Fibre Channel is faster than most LAN media. A larger

number of processors can be connected to the same storage device.

Consolidation: Replacement of multiple independent storage devices by fewer

devices that support capacity sharing. SANs provide the ultimate in scalability,

because software can allow multiple SAN devices to appear as a single pool of

storage accessible to all processors on the SAN. Storage on a SAN can be

managed from a single point of control. Controls over which hosts can see which

storage (called zoning and LUN masking) can be implemented.

Storage Area Network (Cont)


9/99- 9 -

Network Attached Storage


10/99- 10 -

NAS on a network that may be shared with non-storage traffic.

Today, the network is usually an Ethernet LAN, but could be any

network that supports the IP based Protocols like iSCSI.

In contrast to block I/O used by DAS and SANs, NAS I/O

requests are called file I/Os. File I/O is a higher-level type of

request that specifies the file to be accessed, and a number of

bytes to read or write beginning at that offset.

Unlike block I/O, there is no awareness of a disk volume or disk

sectors in a file I/O request.

Network Attached Storage (Cont)


11/99- 11 -

Ease of installation: NAS is generally easier to install and manage than a

SAN. A NAS appliance can usually be installed on an existing LAN/WANnetwork. NAS manufacturers often cite up and running times of 30

minutes or less. Hosts can potentially start to access NAS storage

quickly, without needing disk volume definitions or special devicedrivers. In contrast, SANs take more planning, including design of a Fibre

Channel network and selection/installation of SAN managementsoftware.

Backup: NAS appliances include a snapshot backup facility, to make

backup copies of data onto tape while minimizing application downtime.For SANs, such facilities are available on selected disk systems or in

selected storage management package.

Resource Pooling: NAS allows capacity within the appliance to bepooled. That is, the NAS device is configured as one or more filesystems, each residing on a specified set of disk volumes. All users

accessing the same file system are assigned space within it on demand.

Network Attached Storage (Cont)


12/99- 12 -

Define Storage Area Network (SAN)

Features and benefits of implementing a SAN

Overview of the underlying protocols used within a SAN

SAN Foundation


13/99- 13 -

Logically defined space used by FC

nodes to communicate with each other.

One switch or group of switchesconnected together

Routes traffic between attached devices

Component identifiers: Domain ID

Unique identifier for an FC switchwithin a fabric

Worldwide Name (WWN) Unique 64-bit identifier for an FC

port (either a host port or astorage port)

Basic Structure of SAN

Host

SWITCH

Login

Service

Name

Service

Fabric

Array

Application

O/S

File System


14/99- 14 -

SAN addresses two storage connectivity

problems:

Host-to-storage connectivity: so a host

computer can access and use storage

provisioned to it

Storage-to-storage connectivity: for data

replication between storage arrays

SAN technology uses block-level I/O protocols

Where as NAS uses file-level I/O protocols

The host is presented with raw storage

devices: just as in traditional, direct-attached

storage

Basic Structure of SAN (Cont)

Host

SWITCH

Login

Service

Name

Service

Fabric

Array

Application

O/S

File System


15/99

- 15 -

SAN Connectivity Methods


16/99

- 16 -

There are three basic methods of communication using FibreChannel infrastructure

Point to point (P-to-P)

A direct connection between two devices

Fibre Channel Arbitrated Loop (FC-AL)

A daisy chain connecting two or more devices

Fabric connect (FC-SW) Multiple devices connected via switching technologies

SAN Connectivity Methods (Cont)


17/99

- 17 -

RAID stands for Redundant Array of Independent Disks.

Conceptually, RAID is the use of 2 or more physical disks,to create 1 logical disk, where the physical disks operate intandem to provide greater size and more bandwidth.

RAID has become an indispensable part of any storagesystem today and is the foundation for storagetechnologies.

The use of RAID technology has re-defined the designmethods used for building storage systems.

RAID


18/99

- 18 -

RAID can and will provide excellent I/O performance, whenimplemented with the same care that databaseadministrators have historically taken in designing simpledisk solutions, i.e., separate tables from theircorresponding indexes, if they are accessed in tandem.

On the other hand, it can wreak havoc when implemented ina haphazard fashion.

The two main technical reasons for making the jump toRAID are scalability and high availability in the context ofI/O and system performance.

RAID (Cont)


19/99

- 19 -

Striping : Striping is the process of breaking down data into

pieces and distributing it across multiple disks that support alogical volume Divide, Conquer & Rule.

Mirroring : Mirroring is the process of writing the same data, toanother member of the same volume simultaneously.

Parity : Parity is the term for error checking. Some levels ofRAID, perform calculations when reading and writing data.

Concepts In RAID


20/99

- 20 -

Striping results in a logical volume that is larger and has greater

I/O bandwidth than a single disk. It is purely based on the linearpower of incrementally adding disks to a volume to increase thesize and IO bandwidth of the logical volume. The increase inbandwidth is a result of how read/write operations are done on astriped volume.

A given disk can process a specific number of I/O operations persecond. Anything more than that and the requests start to queueup. By creating a single volume from pieces of data on severaldisks, we can increase the capacity to handle I/O requests in alinear fashion, by combining each disks I/O bandwidth. Now, whenmultiple I/O requests for a file on a striped volume is processed,they can be serviced by multiple drives in the volume, as therequests are sub-divided across several disks. This way all drivesin the striped volume can engage and service multiple I/O requestsin a more efficient manner.

Striping


21/99

- 21 -

Mirroring provides protection for data by writing exactly thesame information to every member in the volume. Additionally,

mirroring can provide enhanced read operations because the

read requests can be serviced from either member of the

volume.

Mirroring


22/99

- 22 -

Some levels of RAID, perform calculations when reading and writing

data. The calculations are primarily done on write operations.However, if one or more disks in a volume are unavailable, thendepending on the level of RAID, even read operations would requireparity operations to rebuild the pieces on the failed disks. Parity isused to determine the write location and validity of each stripe that is

written in a striped volume. Parity is implemented on those levels of

RAID that do not support mirroring.

Parity algorithms contain Error Correction Code (ECC) capabilities,which calculates parity for a given stripe or chunk of data within aRAID volume. The size of a chunk is operating system (O-S) and

hardware specific. The codes generated by the parity algorithm areused to recreate data in the event of disk failure(s). Because thealgorithm can reverse this parity calculation, it can rebuild data, lost

as a result of disk failures.

Parity


23/99

- 23 -

Striping yields better I/O performance.

Mirroring provides data protection.

Parity (when applicable) is a way to check the work.

With these 3 aspects of RAID, we can achieve scalable, protected,

highly available I/O performance.

Putting It All Together


24/99

- 24 -

RAID can be implemented as software-based, where the control

software is usually either bundled with the O-S or in the form ofan add-on.

This type of RAID is also known as host-based RAID. This typeof implementation does impose a small overhead, as it

consumes Memory, I/O bandwidth and CPU on the host where itis implemented.

RAID implemented by hardware in the form of micro-codepresent in dedicated disk controller modules that connect tothe host. These controllers are internal to the host where RAIDis implemented.

The Types Of RAID


25/99

- 25 -

RAID can also be implemented using controllers that are

external to the host where it is implemented. Thisimplementation is bridge-based implementation and is not

preferred, as they incur longer service times for I/O requests.

This is due to the longer I/O paths from the disks to the host.

This type of implementation is usually typical of I/O sub-systems

that are half-Fiber and half-SCSI. It is also common to see

this implementation on storage systems that support multiple

hosts running multiple operating systems.

Hardware-based RAID should be preferred over software-based

or host-based RAID over bridge-based RAID.

The Types Of RAID (Cont)


26/99

- 26 -

RAID levels usually range from 0 to 7.

The differences between the various levels, is based onvarying I/O patterns across the disks.

These I/O patterns by their inherent nature offer different levelsand types of protection and performance characteristics.

The Levels Of RAID


27/99

- 27 -

RAID 0:

This level of RAID is a normal file system with striping, in

which data loss is imminent with any disk failure(s).

In simply worlds, it is data striped across a bunch of disks.

This level provides good read/write performance but norecoverability.

The Levels Of RAID (Cont)


28/99

- 28 -

RAID 1:

In very simple terms this level of RAID provides mirroring and thusfull data redundancy.

This is often called a mirrored disk. In most cases, the volume thatthe operating system sees is made up of two or more disks.However, this is presented to an application or a database as a

single volume. As the system writes to this volume, it writes anexact copy of the data to all members in the volume.

This level of RAID requires twice the amount disk storage ascompared to RAID 0. Additionally, some performance gains can be

reaped from parallel reading of the two mirror members.

RAID 1 doubles the capacity of processing read requests from thevolume when compared to not having mirrored members.

There are no parity calculations involved in this level of RAID.



29/99

- 29 -

RAID 0+1

Stripe First, Then Mirror What You Just Striped. This level of RAIDcombines levels 0 and 1 (striping and mirroring). It also provides goodwrite and read performance and redundancy without the overhead ofparity calculations.

On disk failure(s), no reconstruction of data is required, as the data is

read from the surviving mirror.

This level of RAID is the most common implementation for write-intensiveapplications and is very widely used.

The most common complaint is the cost, since it requires twice as muchspace. To justify this cost, you will have to spend some timeunderstanding the performance requirements and availability needs ofyour systems.

It must be noted here that, the loss of 1 disk of a mirrored member, doesreduce the I/O servicing capacity of the volume by 50%.



30/99

- 30 -

RAID 1+ 0

Mirror First, Then Stripe Over What You Just Mirrored.

This level of RAID has the same functionality as RAID 0+1, but

is better suited for high availability. This is because on the loss

of 1 disk in a mirror member, the entire member of a mirroredvolume does not become unavailable.

It must be noted here that, the loss of 1 disk of a mirrored

member, does not reduce the I/O servicing capacity of the

volume by 50%. This should be preferred method for

configurations that combine striping and mirroring, subject to

hardware limitations.



31/99

- 31 -

RAID 2

This level of RAID incorporates striping, and the

redundancy/protection is provided through parity.

This method requires less disk space compared to RAID 1, but

the need to calculate and write parity, will make writes slower.

This level of RAID was one of the early implementations of

striping with parity using the famous hamming code technique,

but was later replaced by RAID 3, 5 and 7. This level of RAID is

very rarely implemented.



32/99

- 32 -

RAID 3

In this level of RAID, the ECC algorithm calculates parity to provide dataredundancy as in RAID 2, but all of the parity is stored on 1 disk.

The parity for this level of RAID is stored at the bit/byte-level as opposed

to the block/chunk level.

RAID 3 is slowly gaining popularity but is still not very widely used. It isbest suited for data mart/data warehouse applications that support a few

users but require sequential bulk I/O performance (data-transferintensive).

When full table scans and/or index range scans are the norm for a givenapplication and the user population is small, RAID 3 may be just theticket.



33/99

- 33 -

RAID 4

This level of RAID is the same as RAID 3 but with blocklevel parity and is very rarely implemented.



34/99

- 34 -

RAID 5

This is by far one of the most common RAID implementationstoday. In this level of RAID, data redundancy is provided viaparity calculations. Parity is distributed across the number ofdrives configured in the volume.

It results in minimal loss of disk space to parity values, and itprovides good performance on random read operations andlight write operations.

RAID 5 caters better to Input Output Per Second (IOPS) with its

support for concurrently servicing many I/O requests.

It should not be implemented for write-intensive applications,since the continuous process of reading a stripe, calculatingthe new parity and writing the stripe back to disk (with the newparity), will make writes significantly slower.


Th L l Of RAID (C )


35/99

- 35 -

An exception to this rule that requires consideration is when the I/O

sub-system has significant amounts of write cache and the

additional overhead imposed by the ECC algorithms is measured

and confirmed by analysis to be minimal. The definition of

significant is left to the discretion of the reader, but in general a

write cache sized in many gigabytes can be considered significant.

On many systems, however, the performance penalty for write

operations can be expensive even with a significant write cache

depending on the number of writes and the size of each write.

RAID 5 is best suited to read-only applications. Like RAID 3, it is

best suited for data mart/data warehouse applications, but it can

support many application users performing random I/O instead of

sequential I/O.


Th L l Of RAID (C t )


36/99

- 36 -

RAID 6

In this level of RAID, parity is calculated using a more complex

algorithm and redundancy is provided using an advanced multi -dimensional parity method.

RAID 6 stores 2 sets of parity for each block of data and thus makeswrites even slower than RAID 5.

However, on disk failures, RAID 6 facilitates quicker availability of thedrives in the volume (after a disk failure), without incurring the

negative performance impact of re-syncing the drives in the volume.

This level of RAID is very rarely implemented.




37/99

- 37 -

RAID-S

If you are using EMC Storage arrays, then this is yourversion of RAID 3/5. It is well suited to data mart/datawarehouse applications.

This level of RAID should be avoided for write intensive

or high-volume transactional applications for the samereasons as any RAID 5 implementation.

EMC storage solutions are usually configured with largewrite caches, but generally speaking, these write cachesare not large enough to overcome the additional overheadof the parity calculations during writes.




38/99

- 38 -

Auto RAID

With Auto RAID (implemented by HP), the controller along with theintelligence built within the I/O sub-system, dynamically modifies thelevel of RAID on a given disk block to either RAID 0+1 or RAID 5,depending on the near historical nature of the I/O requests on thatblock.

The recent history of I/O patterns on the disk block is maintainedusing the concept of a working set (which is a set of disk blocks).

For obvious reasons, there is one working set each for reads and

writes, and blocks keep migrating back and forth between the twosets, based on the type activity.

A disk block in this context is 64K in size.



39/99

- 39 -

Level of RAID Functionality

RAID 0 Striping, No recoverabili ty, Require read/write performance without Recoverabili ty.

RAID 1 Mirroring, Recoverability, Require write performance

RAID 0+1/1+0 Combination of 0 and 1, Recoverability, Require read and write performance, Very widely

used, 1+0 is better than 0+1 for availability

RAID 2 Early implementation of striping with parity, Uses the Hamming Code Technique for parity

calculations, Was replaced by RAID 3, RAID 5, and RAID 7, Very rarely implemented

RAID 3 Striping with bit/byte-level parity, Dedicated parity disk, Recoverabili ty, Require read

performance for bulk sequential reads, Require data transfer over IOPS, Not widely used but

gaining popularity.

RAID 4 Striping with block-level parity, Dedicated parity disk, recoverability, Very rarely implemented



40/99

- 40 -

RAID 5 Striping with block-level parity, Distributed parity across the number of disks in the volume,

Recoverability, Require read performance for random reads that are small in nature, Require IOPS

over data-transfer, Very widely used

RAID 6 Striping with block-level multi-dimensional parity, Recoverability, Slower writes than RAID 5,Very

rarely implemented

RAID 7 Same as RAID 3, but with better asynchronous capabili ty for reads and writes, Significantly better

overall I/O performance when compared to RAID 3, Significantly more expensive than RAID 3

RAID-S EMCs implementation of RAID 3/5

Auto RAID Hewlett Packards (HP) automatic RAID technology that auto configures the I/O system based on

the nature and type of I/O performed on the disk blocks within the RAID Array.



41/99

- 41 -

Continuous access to information is a must for the smooth

functioning of business operations, as the cost of businessdisruption could be huge, hence Business continuity is an

integrated and enterprise wide process that includes all activities

that a business must perform to mitigate the impact of planned and

unplanned downtime.

Business continuity entails preparing for, responding to, and

recovering from a system outage that adversely affects business

operations. It involves proactive measures, such as business impact

analysis and risk assessments, data protection, security, and

reactive countermeasures, such as disaster recovery and restart, tobe invoked in the event of a failure.

Business Continuity


42/99

- 42 -

Analyzing the business impact of an outage, designing appropriate

solutions to recover from a failure. One or more copies of the originaldata are maintained using any of the following strategies, so that data

can be recovered and business operations can be restarted using an

alternate copy.

Backup and recovery: Backup to tape is the predominant method ofensuring data availability. These days, low-cost, high-capacity disks are

used for backup, which considerably speeds up the backup and

recovery process. The frequency of backup is determined based

frequency of data changes.

Storage array-based replication (local): Data will be replicated within

the same storage array. The replica is used independently for BC

operations. Replicas can also be used for restoring operations if data

corruption occurs.

Business Continuity Planning


43/99

S b d li i (l l)


44/99

- 44 -

Snapview Snapshots

Provide support for consistent on-line backup

Offload backup processing from production hosts

Snapshots can be used testing, decision support scenarios

A successful recovery requires that consistent data written to the

backup media

Storage array-based replication (local):


45/99

- 45 -

Overall highest service level for backup and recovery

Fast sync on first copy, faster syncs on next copy

Fastest restore from Clone

Removes performance impact on production volume

De-coupled from production volume

100% copy of all production data on separate volume

Backup operations scheduled anytime

Offers multiple recovery points

Up to eight Clones against a single source volume

Selectable recovery points in time

Accelerates application recovery

Instantly restore from Clone, no more waiting for tape restore

Snap view Clones


46/99

- 46 -

Snap View Clones and Snap View Snapshots

Each Snap View Clone is a full copy of the source Creating initial Clone requires full sync

Incremental syncs thereafter

Clones may have performance improvements over snapshots in certain situations No Copy On First Write mechanism

Less potential disk contention depending on write activity

Each Clone requires 1x additional disk space

Snapshots Clones

Elements per Source 8 8

Sources per storage system 100 Sources * 50 Clone Groups *

Elements per storage system 800 sessions *

300 snapshots *

100 total images *

S C R


47/99

- 47 -

Adding Clones

Must be exactly equal size to Source LUN

Remove Clones

Cannot be in active sync or reverse-sync process

Termination of Clone Relationship

Renders Source and Clone as independent LUNs

Does not affect data

Source and Clone Relationship

R t R li ti (Mi i / S )


48/99

- 48 -

MirroView

Independent of server, operating system, network applications, and databaseCentralized, simplified management via EMC Navisphere.

Mirrorview software must be loaded on both Primary and Secondary arrays.

Secondary LUN must be the same size as Primary LUN.

Secondary LUN need not be the same RAID type as Primary.

Secondary LUN not accessible to host's.

Mirror must be removed or Secondary promoted to Primary for host to haveaccess. Bi-directional mirroring fully supported.

Remote Replication (Mirrorview / Sancopy)

Primary Secondary

MirrorView Connectivity, Flexibility, and


49/99

- 49 -

Switch attach

FC/IP router . . . . . . . . . . >60km DWDM . . . . . . . . . . . . . . 200km

Optical extender . . . . . . .40 km

Long wave GBIC . . . . . . .10 km

Shortwave GBIC . . . . . . .500 m

Direct attach

CLARiiON to CLARiiON . . 300/500 m

Optical extender . . . . . . . . 10 km

MirrorView Connectivity, Flexibility, and

Distances


50/99


51/99

R t Mi T


52/99

- 52 -

Primary

CLARiiON that serves mirrored primary data to a production host

Secondary

CLARiiON that contains a mirrored secondary copy of primary

data

Mirror Synchronization Mechanism to copy data from primary LUN to a secondary LUN

Mechanism may use fracture log/write intent log to avoid full data

copy

Mirror Fracture

Condition when a secondary is unreachable by the primary

Can be invoked by administrative command

Remote Mirror Terms


53/99

Remote Mirror Functionality


54/99

- 54 -

High Availability

Mirrors resilient to single SP failures

Dual SP protection (primary & secondary copies)

Host I/O allowed to mirror while mirror sync active

Checkpoint of mirror sync progress

Allows sync to continue from last sync checkpoint (if primaryfailure)

Quick recovery of single SP or full failure

Write intent log feature removes full data sync requirement

Mirror I/O can be multiplexed across multiple FC connections

For HA and performance

Remote Mirror Functionality


55/99


56/99


57/99


58/99

How Does SAN Copy Work?


59/99

- 59 -

CLARiiON system acts as a Copy Manager Runs on CLARiiON CX400 through CX700, FC4700/FC4700-2 and later

Can achieve TBs/Hour performance Depends on network infrastructure

Block-level moving/copying of full LUNs Simultaneous push and pull (bidirectional) data movement supported

64 KB granularity for incremental copies

Communicates via World Wide Names Over SAN, LAN or WAN (via FC/IP conversion)

Uses the following devices as source data Snap View Snapshot (full copies only) or Clone Time Finder BCV Idle production LUN

How Does SAN Copy Work?


60/99

SanCopy Topology


61/99

- 61 -

SanCopy Topology

Data

SAN

CLARiiON

SYMMETRIX

LAN

Management

Station

(Navisphere)

Copy

Manager

Data

Data

Data

Data

Data

HOSAGET

Object Copied

is LUN/Volume

Data

SAN Copy: Data Mobility for Business


62/99

- 62 -

Enable better business

decisions Pull data from remote

locations to data center

Gather daily sales records,

Inventory updates

Stop costly data errors

Push data to distributed

locations

Applications, daily pricingupdates, Inventory updates

Reduce operational costs

Centralize data for easier

management

SAN Copy: Data Mobility for Business

Delhi

Atlanta

Mumbai

Corporate DataCenter

Pune

Data

Data

Data

Data


63/99

Managing the Storage


64/99

- 64 -

Management Software

Multipath and Failover

Managing the Storage

SAN Management Tools


65/99

- 65 -

SAN Management Tools

CLARiiON Hardware

FLARE Operating Environment

Navisphere

EMC ControlCenterCLARiiON Based Applications


66/99

CLARiiON Management Options


67/99

- 67 -

There are two CLARiiON management interfaces

CLI (Command Line Interface)

navicli commands can be entered from the command lineand can perform all management functions

GUI (Graphical User Interface)

Navisphere Manager is the graphical interface for allmanagement functions to the CLARiiON array

CLARiiON Management Options


68/99

Navisphere Manager


69/99

- 69 -

Discover

Discovers all managed CLARiiON

systems

Monitor

Show status of storage systems,Storage Processors, disks,snapshots, remote mirrors, and othercomponents

Centralized alerting

Apply and provision

Configure volumes and assignstorage to hosts

Configure snapshots and remote

mirrors

Set system parameters

Customize views via NavisphereOrganizer

Report

Provide extensive performance

statistics via Navisphere Analyzer

Navisphere Manager

Storage Configuration and Provisioning


70/99

- 70 -

Understanding application and server

requirements and planning

configuration

RAID Group is a collection of physical

disks

RAID Protection level is assigned to

all disks within the RAID group

Binding LUNs is the creation of Logical

Units from space within a RAID Group

Storage groups are collections of LUNs

that a host or group of hosts have

access to.

g g g

Step 5 Connect Hosts withStorage Groups

Step 4Add LUNs to

Storage Groups

Step 3 Create Storage Groups

Step 2 Bind LUNs

Step 1 Create RAID Groups

Step 0 - Planning

Creating RAID Groups


71/99

- 71 -

RAID protection levels are setthrough a RAID group

Physical disks part of one RAID group only

Drive types cannot be mixed in the RAID Group

May include disks from any enclosure

RAID types may be mixed in an array

RAID groups may be expanded

Users do not access RAID groups directly

Creating RAID Groups

5 disk RAID-5 group 4 disk RAID-1/0 group


72/99


73/99


74/99


75/99


76/99


77/99


78/99


79/99


80/99


81/99


82/99


83/99


84/99

Path Fault with PowerPath


85/99

- 85 -

If a host adapter, cable,or channeldirector/StorageProcessor fails, thedevice driver returns atimeout to PowerPath

PowerPath responds bytaking the path offlineand re-driving I/Othrough an alternatepath

Subsequent I/Os usesurviving path(s)

Application is unawareof failure

Host Application(s)

HBA HBA

SD SDSD

HBA

Host Bus

Adapter

SCSIDriver

Storage

SERVER

STOR

AGE

InterconnectTopology

SD

HBA

PowerPath


86/99


87/99


88/99


89/99


90/99

I/O with PowerPath Queues in Balance


91/99

- 91 -

PowerPath dynamically

balances workloadacross all available paths

PowerPath will providegreatest performance

improvement inenvironments where theworkload is not balanced

Workloads are seldombalanced

Workloads dynamicallychange

Host Application(s)

Host BusAdapter

SCSIDriver

SD SDSD SD

HBA HBAHBA HBA

Request

Request

Request

Request Request

Request

Request

Request

PowerPath

InterconnectTopology

SE

RVER

STORAGE


92/99

PowerPath Advantages


93/99

- 93 -

Automatic

Dynamic intelligent load management Manages multiple I/O data paths to maximize performance and high

availability

Utilizes multiple data paths to provide greatest efficiency

Nondisruptive Path failover keeps your business in business

Continuous access to information

Online management and configuration

Optimized

Optimizes server and data path utilization by eliminating downtime

Prioritizes bandwidth utilization

Maximizes existing server investment

Business Impact of PowerPath Features


94/99

- 94 -

Optimized performance and high

availability; no applicationdisruption

Consistent and improved servicelevels

Improved manageabilitysavestime, reduces maintenance cost

Optimized data managementthrough user-selectable storageallocation policies

Automated information utilization;optimized data movement

Automatic path failover and recovery

Dynamic load balancing of I/O

Online configuration andmanagement

Policy-based management

Automated server-to-storage I/Omanagement


95/99

PowerPath Interoperability


96/99

- 96 -

PowerPath Verses Other Products


97/99

- 97 -

Veritas DMP

Provides failover and limited load balancing capability

SUN Alternate Pathing

Failover only

HP PVlinks

Failover only

Windows MPIO

Failover only


98/99


99/99

Documents

Storage Fundamentals