Dell PowerVault 221S User Manual

;

A Reference Guide for Optimizing Dell™ SCSI Solutions

Authored By:

Dell SCSI Storage Solution Team

November 17, 2005 rev A02

____________________

Information in this document is subject to change without notice.

© Copyright 2005 Dell Inc. All rights reserved.

Reproduction in any manner whatsoever without the written permission of Dell Inc. is strictly
forbidden.

THIS WHITE PAPER IS FOR INFORMATIONAL PURPOSES ONLY, AND MAY CONTAIN
TYPOGRAPHICAL ERRORS AND TECHNICAL INACCURACIES. THE CONTENT IS
PROVIDED AS IS, WITHOUT EXPRESS OR IMPLIED WARRANTIES OF ANY KIND.

Dell, the Dell Logo, PowerVault, PowerEdge and OpenManage are trademarks of Dell Inc.
Microsoft is a registered trademark of Microsoft Corporation. Other trademarks and trade names

may be used in this document to refer to either the entities claiming the marks and names or their
products. Dell disclaims proprietary interest in the marks and names of others.

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

TABLE OF CONTENTS

A REFERENCE GUIDE FOR OPTIMIZING DELL SCSI SOLUTIONS............................................................... 3

1. INTRODUCTION ................................................................................................................................................... 3

2. STORAGE APPLICATIONS AND COMPONENTS.................................................................................................... 4

Identify customer usage model and needs............................................................................................................. 4

Picking a solution based on the application profile.............................................................................................. 4

General RAID Configuration Guidelines .............................................................................................................................. 5

Application specific guidelines............................................................................................................................... 6

E-Mail Servers......................................................................................................................................................................... 6

Database Servers .................................................................................................................................................................... 7

Web Servers............................................................................................................................................................................. 8

Online Transaction Processing Servers................................................................................................................................. 9

File Servers............................................................................................................................................................................ 10

Streaming Media Servers...................................................................................................................................................... 12

Additional configuration factors.......................................................................................................................... 13

Hot Spares ............................................................................................................................................................................. 13

Clustering Considerations.................................................................................................................................................... 13

Considerations for Optimal RAID Performance................................................................................................................. 13

3. DELL’S PRODUCT OFFERING ............................................................................................................................ 14

PowerVault™ 220S & 221S Storage Enclosure.................................................................................................14

PERC – PowerEdge™ RAID Controllers ..........................................................................................................14

SCSI Cabling.........................................................................................................................................................15

SCSI Hard Drive Support..................................................................................................................................... 16

3rd Party Hardware support................................................................................................................................. 16

Server Support ......................................................................................................................................................16

4. PV22XS ENCLOSURE OVERVIEW .................................................................................................................... 19

Joined Bus Topology ............................................................................................................................................ 19

Split Bus Topology................................................................................................................................................ 20

Cluster Topology ..................................................................................................................................................21

PV22xS – Common Features and Capabilities................................................................................................... 22

SCSI ID Assignment.............................................................................................................................................................. 22

Multiple PV22xS Enclosures on a Single Host System .......................................................................................................23

Mixing EMM Speeds............................................................................................................................................................. 23

Mixing HDD Capacities ....................................................................................................................................................... 24

Mixing HDD Spindle Speeds................................................................................................................................................ 24

SCSI Enclosure Services (SES) ............................................................................................................................................24

Warm and Hot Pluggable Features ..................................................................................................................................... 24

External SCSI Cables............................................................................................................................................................ 25

External SCSI Connector Types........................................................................................................................................... 25

PV22xS Configuration Topology and Application Attributes............................................................................................. 26

5. SCSI HOST BUS ADAPTERS / SCSI RAID....................................................................................................... 29

RAID: Hardware Vs Software .............................................................................................................................29

Considerations to Implement a RAID Solution................................................................................................... 29

Controller Family.................................................................................................................................................................. 30

Data Protection Strategy ......................................................................................................................................................32

High Availability Needs – Clustering................................................................................................................................... 33

Heterogeneous vs. Homogeneous environments................................................................................................................. 33

Interface Channel Options – Internal/External/Multi channel device configuration........................................................ 34

Capacity Needs – Virtual Disk Size...................................................................................................................................... 35

Additional RAID information............................................................................................................................... 35

6. SUMMARY AND CONCLUSIONS ........................................................................................................................ 39

7. APPENDIX – A: REFERENCES............................................................................................................................ 41

8. APPENDIX – B: GLOSSARY ............................................................................................................................... 42

9. APPENDIX – C: ABOUT THE DELL SCSI STORAGE SOLUTION TEAM .............................................................43

Authors .................................................................................................................................................................. 43

10. APPENDIX – D: REVISION HISTORY ............................................................................................................. 44

PAGE 2 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

A Reference Guide for Optimizing Dell SCSI Solutions

1. Introduction

Enterprise storage applications involve interconnection of multiple components – including
both software and hardware aspects which would complete a storage solution. Any given
storage application may be supported by one or more storage solutions. Thus, it becomes
essential to identify an optimum storage solution, one which would be ideal for a given
application.

Selection of storage components not only impacts the effectiveness of the solution for an
application but also would impact the performance of Enterprise IT infrastructure which
relies on that storage. Key storage components which impact a storage solution can be
categorized as:

• Physical Storage Device Technology – PowerVault™ Direct Attached Storage
PV22xS family

• Interconnect Technology – Ultra160/Ultra320 SCSI

• Host System Interface Technology – non-RAID (HBA) and RAID controllers

(PERC)

• Management Capabilities – OpenManage™ Storage Management (OMSM) or
Array Manager

While selection of storage components is essential for building a storage solution, on the
other hand it is also critical to understand the requirements for a given application. In
general, effectiveness of all the applications is driven by the following key storage
attributes and the usage model for the application, which should then determine the
relative importance for each of these attributes:

• Reliability – Storage solution provides reliable access to data

• Availability – Data from storage devices readily available

• Serviceability – In case of failure of storage components, solution should allow

easy access to assemblies/components and be easy to service

• Redundancy – Solution provides redundancy for both the data storage, as well as
for hardware management capabilities

• Performance

o Bandwidth Utilization – Data Throughput/Available Bandwidth
o I/Os per second (IOPS)

• Data Protection – Data is protected against catastrophic failures on a storage or
system component

• Scalability – Storage capacity can be expanded as applications necessitate

• Cost – driven by RAID and other hardware configurations.

An optimum storage solution would be something comprised of storage components that
would address all the key attributes for a given application, keeping their relative
importance in mind. This paper presents a study of all storage components from
technological to functional, as well as behavioral differences, to help define an optimum
storage solution. In this paper, all the storage components are measured against the
attributes listed above which impact the overall effectiveness of an application.

PAGE 3 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

Understanding each of the storage components with respect to these attributes will help in
making intelligent decisions to determine an optimal configuration for a given application.

The paper presumes that the reader has a basic understanding of different levels of RAID
and interconnects technologies – Parallel SCSI (SPI-4), PCI vs. PCI-X vs. PCI-e. This
paper presents a study of all storage components, as identified earlier, against various
storage applications and different operating environments; however, specific behavior of
each of the storage applications and management functions is beyond the scope of this
document. In this document we present storage solution from Parallel SCSI perspective,
specifically for Dell’s Direct Attached Storage (DAS) products. Other storage technologies
such as SAS, SATA, or Fibre Channel, as well as topologies such as Network Attached
Storage (NAS) or Storage Area Network (SAN) are outside the scope of this document. All
storage applications referred to herein are treated independent of any specific host system
or server.

2. Storage Applications and Components

Identify customer usage model and needs

In order to select the correct storage solution for any given situation, it is important to
understand what the application and user requirements will be for that solution. A good
starting point is to have an understanding of basic storage profiling considerations.

• Table 2-1: Storage Profiling Considerations

Characteristic Values Description

Performance is the overall ability of the solution to read
and write data to disk. The performance requirements
are usually determined by the type of application being
utilized. Different applications have different
performance requirements. For example, a database
or e-mail server has mostly random disk access
operations while a streaming media server would have
mostly sequential disk access.

Performance

• Bandwidth (MB per
sec.)

• I/O size (KB/MB)

• I/O Profile (read/write

and random/sequential
access mix)

• Latency

Storage capacity is the current amount of storage

Storage Capacity

Needs

Gigabytes

space required by the application and user data. For
example, e-mail storage for 100 users would require
much less storage capacity than an e-mail store for
1000 users.
Storage growth is the expected increase in the amount

Storage Growth

Rate

Percent increase per

year

of capacity that will be required as usage of the server
increases. This is usually measured in the expected
number of users or clients accessing the server

Criticality defines the impact to mission critical business

Criticality

Low, Medium, High,

Very High

needs if the storage is offline. This characteristic is
important for choosing the right RAID level, and
determining if clustering is needed.

Picking a solution based on the application profile

One of the most important factors to consider when selecting a storage solution is the type
of application that will be utilizing the storage solution. This defines the overall purpose of

PAGE 4 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

the server/storage solution and will determine what RAID configuration will be the most
optimal for the application. Table 2-2 outlines the recommended RAID solution based on
the usage of the server.

• Table 2-2: Application Vs RAID configurations

Application

Email

Database

Web

On-line
transaction

File ­Archival

File - User
File stores

Streaming
Media

RAID Level

Concatenated 0 1 10 5 50

Recommended Not Recommended Possible

General RAID Configuration Guidelines

RAID 0

RAID 0 is generally not a recommended solution due to lack of redundancy and data
protection. However, it may be utilized in situations where these are not required and
maximum storage capacity and performances are essential.

RAID 1

A RAID 1 solution, while not ideal for most servers, could be utilized for small workgroup
servers or servers with low storage capacity and growth requirements. Any servers
beyond a small workgroup server are likely to need more storage and consequently, a
more efficient RAID solution.

RAID 10

RAID 10 has good I/O performance, excellent availability and redundancy. The biggest
drawback is the higher relative cost for significantly lower storage capacity. This RAID
configuration should be used in situations where the maximum availability, redundancy
and performance are the priority considerations.

PAGE 5 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

RAID 5

RAID 5 has good I/O performance, data protection and relative cost with excellent storage
capacity. RAID 5 should be used in situations where maximum storage capacity is
required along with a moderate amount of data protection.

RAID 50

A balance between RAID 5 and RAID 10, this solution offers good I/O performance,
availability and good storage capacity. This configuration offers slightly higher
performance than RAID 5, but at a slightly higher cost and reduced storage capacity. It
does however provide greater storage capacity than a RAID 10 solution with a decrease in
performance and data protection. This should be used in situations where greater
redundancy and data protection is required as well as reasonable storage capacity.

Note: For more details on RAID configurations refer to the Dell™ ‘Getting Started with
RAID’ document (http://support.dell.com/support/edocs/storage/RAID/RAIDbk0.pdf

).

Concatenated Container

This solution is not recommended due to lack of data protection and redundancy and no
performance gain. However for non-critical implementations that require high scalability
this type of solution will provide more ease of use than other RAID configurations.

Application specific guidelines

E-Mail Servers

The storage requirements for e-mail servers can vary depending on the size, the amount,
and type of users. While small departmental e-mail servers may work well with a small
amount of storage and limited features; large corporate e-mail servers normally require
greater storage capacity, very high availability, performance and scalability. IO profiles will
vary depending on the number of users and type of mail and attachments sent.

• Table 2-3: Email Server General I/O profile

I/O Profile

(Read/Write)

I/O Profile

(Sequential/Random)

Bandwidth

I/O

Size

Latency

Sensitivity

Growth

Rate

Critical

ity

60/40 Random Heavy 4k High High High

• Table 2-4: Email Sever RAID Guidelines

Application

E-Mail

Concatenated 0 1 10 5 50

RAID Level

Recommended Not Recommended Possible

Recommended:

PAGE 6 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

• RAID 10 – Recommended for enterprise mission critical e-mail servers where
availability and redundancy and/or performance are the highest priority and
reduce storage capacity is acceptable.

• RAID 5 – Recommended for e-mail servers that require maximum storage
capacity and only base data protection and performance.

• RAID 50 – Recommended for those solutions that require greater redundancy
and data protection and a balance between storage capacity and performance.

Possible:

• RAID 1 – Possible solution for small e-mail servers which do not require high
storage capacity.

Not recommended:

• RAID 0, Concatenated – Not recommended due to lack of redundancy and data
protection.

Note: While these configurations are not recommended, they can be configured
and utilized.

Database Servers

Database servers can range from simple workgroup databases like Microsoft® Access
with a few hundred users to mission critical enterprise databases like Oracle or SQL
Server with thousands of users. Database applications will always benefit from some data
protection while other requirements such as performance and availability will vary. As a
general rule, the more mission critical the database is, the more data protection is
required. Additionally, the performance requirements increase relative to the number of
users accessing the database.

• Table 2-5: Database General IO profile

I/O Profile

(Read/Write)

I/O Profile

(Sequential/Random)

Bandwidth IO Size

Latency

Sensitivity

Growth

Rate

Criticality

70/30 Random Heavy 8K High High High

• Table 2-6: Database sever RAID Guidelines

Application

Database

Concatenated 0 1 10 5 50

RAID Level

Recommended Not Recommended Possible

Recommended:

• RAID 10 – Recommended for enterprise mission critical database solutions
where availability and redundancy and/or performance are the highest priority.

PAGE 7 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

• RAID 5 – Recommended for databases that require maximum storage capacity
and only base data protection and performance.

• RAID 50 – Recommended for those solutions that require a balance between
storage capacity and performance.

Possible:

• RAID 1 – Possible solution for small databases which do not require high storage
capacity.

Not recommended:

• RAID0, Concatenated - Not recommended due to lack of redundancy and data
protection

Note: While these configurations are not recommended, they can be configured
and utilized.

Web Servers

Web severs are usually high traffic systems where read operations are the most common
disk activity as web pages are requested by users. They can be moderate intranet sites
with minimal traffic internal company traffic or global internet portal sites that receive
hundreds of thousands of users daily. Performance is generally the main concern with
redundancy and data protection being less a concern. For large web sites, usually there
are several web servers responding to client requests and as a result the need for
redundancy and data protection

is not as important. Also web servers may not require as
much scalability or disk capacity since website content is relatively static and changes are
usually minor content modifications.

• Table 2-7: Web General IO profile

I/O Profile

(Read/Write)

I/O Profile

(Sequential/Random)

Bandwidth IO Size

Latency

Sensitivity

Growth

Rate

Criticality

95/5 Random Moderate < 64K Moderate Low High

• Table 2-8: Web RAID Guidelines

Application

Database

Concatenated 0 1 10 5 50

RAID Level

Recommended Not Recommended Possible

Recommended:

• RAID 10 – Recommended for enterprise web server solutions where availability
and redundancy and performance are the highest priority, usually stand-alone
portal sites that is critical to the organization’s business.

PAGE 8 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

• RAID 5 – Recommended for web servers that require maximum storage capacity
and only base data protection and performance.

• RAID 50 – Recommended for those solutions that require a balance between
storage capacity and performance.

• RAID 0 – Recommended for solutions where web server will be part of a group of
web servers that service a large internet portal and performance is the highest
concern. In this situation, availability and redundancy is handled by the cluster
group and does not need to happen at the disk level.

• RAID 1 – Good solution for small websites which do not require high storage
capacity.

Not recommended:

• Concatenated - This solution is not recommended due to lack of redundancy and
data protection

Note: While this configuration is not recommended, it can be configured and utilized.

Online Transaction Processing Servers

Online Transaction Processing (OLTP) oriented servers are used in a number of
industries for the entry and retrieval of transactions for example, OLTP is common in
banking, airlines, mail-order, and supermarkets. They are generally mission critical
servers that require the maximum availability and redundancy possible.

• Table 2-9: OLTP General IO profile

I/O Profile

(Read/Write)

I/O Profile

(Sequential/Random)

Bandwidth IO Size

Latency

Sensitivity

Growth

Rate

Criticality

80/20 Random Moderate 2k-8k Moderate Low High

• Table 2-10: OLTP RAID Guidelines

Application

OLTP

Concatenated 0 1 10 5 50

RAID Level

Recommended Not Recommended Possible

Recommended:

• RAID 10 – Since OLTP systems are very critical for most business, this RAID

configuration is highly recommended due to the high availability and
redundancy and good performance.

• RAID 5 – Recommended for OLTP servers that require maximum storage

capacity and only base data protection and performance.

PAGE 9 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

• RAID 50 – Recommended for those solutions that require a balance between

storage capacity and performance.

Possible:

• RAID 1 – Possible solution for situations which do not require high storage

capacity.

Not recommended:

• RAID 0, Concatenated - These are not recommended due to lack of

redundancy and data protection.

Note: While these configurations are not recommended, they can be configured and

utilized.

File Servers

File servers can be archival long term storage repositories or more dynamic user file
storage where files are changed, added and deleted on a daily basis. They can range
from workgroup to company level in their scope. A key factor for file servers is storage
capacity as users add more files. File servers are generally not mission critical systems so
reduced availability and redundancy or none at all is acceptable as the data is usually
backed up and can be restored in a matter of hours.

Archival file server characteristics and recommendations

• Table 2-11 Archival File Server General I/O profile

I/O Profile

(Read/Write)

I/O Profile

(Sequential/Random)

Bandwidth IO Size

Latency

Sensitivity

Growth

Rate

Criticality

90/10 Sequential Moderate >64K High Varies Low

• Table 2-12: Archival File Server RAID Guidelines

Application

File -

Archival

Concatenated 0 1 10 5 50

RAID Level

Recommended Not Recommended Possible

Recommended:

• RAID 10 – Recommended due to the high availability and redundancy and

good performance.

• RAID 5 – Recommended for file servers that require maximum storage

capacity and only base data protection and performance.

• RAID 50 – Recommended for those solutions that require a balance between

storage capacity and performance.

Possible:

PAGE 10 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

• RAID 1 – Possible solution in situations which do not require high storage

capacity.

Not recommended:

• RAID 0, Concatenated - These are not recommended due to lack of

redundancy and data protection. Since long term archival storage is indented
in some ways to be a backup of important files, it is not a recommended
solution.

Note: While these configurations are not recommended, they can be configured and

utilized.

User file store characteristics and recommendations

• Table 2-13 User Store File Server General IO profile

I/O Profile

(Read/Write)

I/O Profile

(Sequential/Random)

Bandwidth IO Size

Latency

Sensitivity

Growth

Rate

Criticality

80/20 Sequential Heavy >64K High Varies Moderate

• Table 2-14: User Store File Server RAID Guidelines

Application

File – User

file stores

Concatenated 0 1 10 5 50

RAID Level

Recommended Not Recommended Possible

Recommended:

• RAID 5 – Recommended for file servers that require maximum storage

capacity and only base data protection and performance.

• RAID 50 – Recommended for those solutions that require a balance between

storage capacity and performance.

Possible:

• RAID 1 – Possible solution situations which do not require high storage

capacity.

• RAID 10 – Possible solution for situations that require greater availability and

redundancy but not as much disk capacity.

• RAID 0 – While this solution is not fault tolerant, RAID 0 may be useful for file

servers that require better performance and maximum storage capacity, for
example, where backups are expected to maintain data in the event of a failure.

• Concatenated – This solution could be use for situations that require maximum

scalability as drives could be added as storage needs grow. For file servers

PAGE 11 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

that can afford some downtime and rely on backups to restore data, this could
be a solution.

Streaming Media Servers

Streaming media servers are systems that provide web casting, video conferencing,
internet entertainment like TV or radio, and multimedia content oriented services. These
are systems that generally require a balance between storage capacity, availability,
redundancy and performance. Like web servers, they may also be part of a group of
systems that work together to provide content.

• Table 2-15 Database General IO profile

I/O Profile

(Read/Write)

I/O Profile

(Sequential/Random)

Bandwidth IO Size

Latency

Sensitivity

Growth

Rate

Criticality

98/2 Sequential Heavy >64K High High Varies

• Table 2-16: Database sever RAID Guidelines

Application

Streaming

Media

Concatenated 0 1 10 5 50

RAID Level

Recommended Not Recommended Possible

Recommended:

• RAID 5 – Recommended for streaming media servers that require maximum

storage capacity and only base data protection and performance. This would
be recommended RAID configuration for stand alone streaming media servers.

• RAID 50 – Recommended for those solutions that require a balance between

storage capacity and performance. This solution would provide greater
availability and redundancy but less storage capacity.

• RAID 0 – While this solution is not redundant, this may be useful for streaming

media servers that require better performance and are members of a cluster of
servers where availability and redundancy is handled at a higher level.

Possible:

• RAID 1, RAID 10 – Possible solutions in situations which do not require high

storage capacity. In general servers utilized for streaming media require large
amounts of storage capacity for content especially uncompressed video.

Not Recommended:

• Concatenated – This solution is not recommended due to lack of redundancy

and limited performance.

Note: While these configurations are not recommended, they can be configured and
utilized.

PAGE 12 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

Additional configuration factors

Hot Spares

Hot Spare functionality provides extra security availability and redundancy by replacing a
failed drive, allowing the rebuild of the degraded array to begin immediately. Both the
Adaptec and LSI based PERC controllers’ support this feature, but a simple SCSI
controller would not. See the controller user’s guide for details and limitations. Use of Hot
Spares is recommended whenever possible.

Clustering Considerations

In a cluster environment the cache being disabled on the controller poses a significant
performance impact to RAID 5 and RAID 50 configurations. If the system is to be clustered
RAID 10 is the optimal RAID level.

• Table 2-17: Cluster Solution Vs RAID Levels

RAID Level

Cluster

Concatenated 0 1 10 5 50

Solutions

Considerations for Optimal RAID Performance

The PowerVault™ 22xS is capable of offering increased performance by allowing multiple
channels to connect to the storage enclosure. This ability can increase the performance
of most configurations by allowing a multi channel RAID controller to increase the
bandwidth of the SCSI bus. For more information see Table 4-2: PV22xS Configuration
Topologies vs RAID Level Performance.

Clustering Multi Channel Relative Cost

Concatenated drives

RAID 0
RAID 1

RAID 10

RAID 5

RAID 50

Not Recommended Recommended Possible

• Table 2-1 2-7: RAID Level Benefit (Generalized)

None None None
None Excellent None

Moderate Excellent Excellent

Excellent Excellent Excellent
Moderate Moderate Moderate
Moderate Moderate Moderate

Capacity

Excellent
Excellent

None

None
Moderate
Moderate

Scalability

Excellent

None
None

Excellent
Moderate
Moderate

PAGE 13 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

3. Dell’s Product Offering

PowerVault™ 220S & 221S Storage Enclosure

Dell™ PowerVault 220xS systems are flexible, external SCSI expansion enclosures
designed to support multiple Dell storage environments and RAID configurations. Each
system offers maximized drive-spindle count, hot-plug hard drives, optional redundant
power, redundant cooling, rack mount capability, systems management features, and a
modular design for easy upgrades. Most major components, including hard drives and
power supply/cooling modules are hot-pluggable and can be removed and replaced
easily.

The PV22xS is available as two model types: the PV220S which fits within an industry
standard 19 inch rack, and the PV221S which is a free standing floor tower. Both models
leverage the same base mechanical frame, but differ with respect to the mounting
hardware and front bezel features. Dell offers a conversion kit to change from a PV220S
to a PV221S. Basic features and options common to both models are shown in Figure
3-1.

Performance



U320 SCSI speed



dual channel PCI-Express PERC 4

U320 controller

Capacity



4.2TB with 300GB hard drives



Up to 32 disks per virtual disk

Scalability



Up to eight enclosures per server



Scales up to 32TB per server

Manageability



Robust RAID management



Fully integrated solution

Affordability



Lower cost than Fibre Channel



Low cost non-redundant configs

PowerVault 220S/221S Storage Arrays

PV 220S Rack

SCSI Storage Array

FEATURES

Up to 14 1” SCSI U160 SCSI hard disk drives




73GB, 146 & 300GB 10K RPM or 18GB, 36GB & 73GB 15K RPM
disks



Supports RAID 0, 1, 5, 10, 50 with PERC 3 or PERC 4 RAID cards



3+1 Fan Redundancy, optional redundant power supply



Supports cables up to 12 meters from host server



Configurable in joined bus, split bus or cluster mode



Leverages PERC 3 U160 or PERC 4 RAID controller card



Allows 32 hard disk drive per virtual disk



Up to forty (40) virtual disks per controller



Dual channel and Quad Channel



Host clustering support (Microsoft Cluster Support)



Externally accessible Ultra3 (Ultra160) or Ultra 320 (U320) SCSI



3U form factor for rack dense environments

PV 221S Tower

SCSI Storage Array

Figure 3-1 PV22xS Enclosure Features and Options

PERC – PowerEdge™ RAID Controllers

The PowerEdge RAID Controller (PERC) is a high-performance, intelligent peripheral
component interconnect (PCI) and PCI-Express to Small Computer System Interface

PAGE 14 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

(SCSI) host adapter with RAID control capabilities. It provides reliable fault-tolerant disk
subsystem management and is an ideal RAID solution for internal storage in Dell's
PowerEdge enterprise systems. The RAID controller offers a cost-effective way to
implement RAID in a server.

Dell is currently offering its fourth generation of RAID controllers, the PERC 4e family,
supporting the industry standard Ultra320 SCSI storage interconnect. PERC 4e RAID
controllers support the PCI-Express host interconnects for maximum bandwidth
connectivity & expandability. The PERC 4e family of controllers includes the PERC 4e/SI
and PERC 4e/DI for RAID On Motherboard solutions and the PERC 4e/DC add-in HBA.
The following matrix describes the benefits of each solution.

Previous generations of PERC have included the PERC 2, PERC 3 and PERC 4 families.
The following PERC solutions support the PowerVault 220S and 221S External SCSI
Storage solution:

• Table 3-1: PERC Solution supported by PV22xS

Maximum

Card

Speed

80 MB/s PERC 2/DC

PERC 2/QC

160 MB/s 39160 HBA
PERC 3/DC
PERC 3/QC
320 MB/s PE 2600 ROMB (PERC 4/Di on this server)

PE 1750 ROMB (PERC 4/Di on this server)
PERC 4/SC
PE 2800 ROMB (PERC 4e/Di on this server)
PE 2850 ROMB (PERC 4e/Di on this server)
PE 6800 ROMB (PERC 4e/Di on this server)
PE6850 ROMB (PERC 4e/Di on this server)
PERC 4/DC (PCI, dual channel U320 RAID add-in HBA)
PERC 4e/DC (PCI-Express, dual channel U320 RAID add-

in HBA)

SCSI Cabling

HBA and PERC controllers support a low-voltage differential (LVD) SCSI bus. LVD
enables cables up to 12 meters long. Throughput on each SCSI channel can be as high
as 320 MB/sec.

PAGE 15 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

• Table 3-2: PERCs and Cable Type

PV22xS Controller/Cable Matrix

Controller type Cable type

PERC Add-in Cards

PERC 2/DC Wide HD to Slim VHDCI
PERC 2/QC

Wide HD to Slim VHDCI

PERC 3/QC Wide HD to Slim VHDCI
PERC 3/DCL Wide HD to Slim VHDCI

PERC 3/Di Embedded

Wide HD to Wide HD on PE2600, PE2500, and 4400
Slim VHDCI to Wide HD on PE1650, PE1750

PERC 4/DC Wide HD to Slim VHDCI
PERC 4/Di Embedded

Wide HD to Wide HD on PE2600

PERC 4e/DC Wide HD to Slim VHDCI
PERC 4e/Di Embedded Wide HD to Wide HD

SCSI HBAs

Adaptec 3960D (39160) Wide HD to Wide HD

SCSI Hard Drive Support

PowerVault disk-based storage solutions (such as the PowerVault 220S) support the
latest SCSI hard drives, currently specified as Ultra 320 SCSI. SCSI solutions from Dell
support both 10,000 RPM and 15,000 RPM spindle speeds and several variations of
capacities to allow customers the maximum degree of flexibility. For the latest information
of support for specific hard drives, please visit the Dell Storage web site at:

http://www1.us.dell.com/content/products/category.aspx/storage

rd

Party Hardware support

3

.

Dell development teams spend a significant amount of time testing and qualifying SCSI
hard drives and PERC controllers with PowerVault 22xS, and PERC controllers (as well
as PowerEdge servers), to ensure the highest quality products. As a result of this testing,
hard drives purchased direct from Dell can include upgrades not generally available to the
rest of the industry. As such, Dell supports only Dell tested and qualified hardware. In
order to help ensure the best possible customer experience, Dell highly recommends
purchasing all hardware direct from Dell for the PowerVault 22xS, PowerEdge servers or
any other Dell products. All hardware (disk drives, RAID controllers, cables, power
supplies) purchased direct from Dell carries the Dell limited warranty
compatibility. Materials purchased from 3

rd

party vendors for use in Dell systems are not

1

and helps ensure full

protected by a Dell warranty and may not deliver an acceptable customer experience.

Server Support

Dell PowerVault SCSI Solutions support all PowerEdge, PowerVault NAS and many
PowerEdge SC servers. This adds tremendous value by delivering a single integrated
Enterprise solution that leverages industry standard hardware platforms and a common
management framework for customers’ server and storage solutions.

1

For a copy of our guarantees or limited warranties, please write Dell USA L.P., Attn: Warranties, One Dell Way, Round Rock, TX 78682.

For more information, visit www.dell.com/warranty

PAGE 16 11/17/2005

.

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

In addition, many Dell servers & storage solutions offer cross generational support to
ensure customers have a seamless experience when adopting new technologies. The
following compatibility matrix indicates the cross-generational support Dell offers for its
SCSI Solutions:

• Table 3-3: PERC 4 Support Matrix (6G, 7G & 8G servers)

Y = Supported, N = Not Supported, # = maximum # of factory installed cards
FI = Factory install; CK = Customer Kit

System 4/Di (ROMB) 4/Si (ROMB) 4/IM 4e/Di (ROMB) 4e/Si (ROMB)

FI FI FI FI CK FI CK FI CK FI FI

PowerEdge Servers

PE 400SC N N NNNNNNN N N
PE 420SC N N NNNNNNN N N
PE 300 N N NNNNNNN N N
PE 350 N N NNNNNNN N N
PE 500SC N N NNNNNNN N N
PE 600SC N N NNNNNNN N N
PE 650 N N N1Y1YNN N N
PE 700 N N N 1 YNNNN N N
PE 750 N N N1Y1YNN N N
PE 800 N N NNN 1 YNN N N
PE 1400 N N NNNNNNN N N
PE 1400SC N N N 1 Y N N N N N N
PE 1420SC N N NNNNNNN N N
PE 1500 N N N 1 Y N N N N N N
PE 1550 N N NNNNNNN N N
PE 1600SC N N NNNNNNN N N
PE 1650 N N N N N 1 Y N N N N
PE 1655MC N N NNNNNNN N N
PE 1855MC N N NNNNNNN N N
PE 1750 Y N N N N 1 Y N N N N
PE 1800 N N N1Y1Y1Y Y N
PE 1850 N Y N1Y1Y1Y N Y
PE 2400 N N NNNNNNN N N
PE 2450 N N NNNNNNN N N
PE 2500 N N N N N 2 Y N N N N
PE 2550 N N N N N 2 Y N N N N
PE 2600 Y N N N N 2 Y N N N N
PE 2650 N N N N N 2 Y N N N N
PE 2800 1 N N N N 2 Y 2 Y Y N
PE 2850 1 N N N N 2 Y 2 Y Y N
PE 2855 N N YNNNNNN N N
PE 3250 N N N 1 Y 2 Y N N N N
PE 7250 N N NNNNNNN N N
PE 4400 N N N N N 2 Y N N N N
PE 4600 N N N N N 2 Y N N N N
PE 6400 N N N N N 3 Y N N N N
PE 6450 N N N N N 3 Y N N N N
PE 6600 N N N N N 3 Y N N N N
PE 6650 N N N N N 3 Y N N N N
PE 6800 1 N N N N 3 Y 3 Y Y N
PE 6850 1 N N N N 3 Y 3 Y Y N
PE 7150 N N NNNNNNN N N
PE 8250 N N N N N 3 Y N N N N

PowerVault NAS N N NNNNNNN N N

PV 725N N N NNNNNNN N N
PV 735N N N NNNNNNN N N
PV 745N N N N 1 Y 1 Y N N N N
PV 750N N N NNNNNNN N N
PV 755N N N NNNNNNN N N
PV 770N Y N N N N 2 Y N N N N
PV 775N N N N N N 2 Y N N N N

PV220 Support

PV22xS Y Y NYYYYYY Y Y

NNNNN NNN N

4/SC 4/DC 4e/DC

PAGE 17 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

• Table 3-4: PERC 2 & PERC 3 Support Matrix (4G & 5G servers)

Y = Supported, N = Not Supported, # = maximum # of factory installed cards
FI = Factory install; CK = Customer Kit

System

PowerEdge Server Support

PE 500SC NNNNNN1 YNNNNNN
PE 600SC NNNNNN1 YNNNNNN
PE 1400 NNNNNN1 Y1 YNNNN
PE 1400SC NNNNNN1 Y 1 YNNNN
PE 1500 NNNNNN1 YNNNNNN
PE 1550 NNNNNNNN1 Y1 YNN
PE 1600SC NNNNNN1 YNNNNNN
PE 1650 N N N N 1 Y N N N N 1 Y N N
PE 2400 NN 1YNNNNNN 2 Y 2 Y
PE 2450 NN 1YNNNNNN 2 Y 2 Y
PE2500 N N N N 1 Y N N N N 2 Y 2 Y
PE 2550 N N N N 1 Y N N N N 2 Y 2 Y
PE 2600 NNNNNNNNNN2 Y2 Y
PE 2650 N N N N 1 Y N N N N 2 Y 2 Y
PE 3250 NNNNNNNNNN2 YNN
PE 7250 NNNNNNNNNNNNNN
PE 4400 NN 1YNNNNNN 2 Y 2 Y
PE 4600 N N N N 1 Y N N N N 2 Y 2 Y
PE 6400 NN 1YNNNNNN 3 Y 2 Y
PE 6450 NN 1YNNNNNN 3 Y 2 Y
PE 6600 NNNNNNNNNN3 Y2 Y
PE 6650 NNNNNNNNNN3 Y2 Y
PE 7150 NNNNNNNNNN3 Y2 Y
PE 8250 1Y1YNNNNNN3Y2Y

3/DI2/DC2/SC 3/QC3/SC 3/DCL 3/DC

FI CK FI CK FI CK FI CK

PowerVault NAS Support

PV 725N NNNNNNNNNNNNNN
PV 735N NNNNNNNNNN 1 Y2 Y
PV 750N NNNNNNNNNN 2 Y2 Y
PV 755N NNNNNNNNNN 1 Y2 Y
PV 770N NNNNNNNNNN 2 Y2 Y
PV 775N NNNNNNNNNN 1 Y2 Y

PowerVault Enclosure Support

PV20xS YYYYNNNNYYYYYY
PV21xS YYYYNNNNYYYYYY
PV22xS NNNNNNNNYYYYY Y

PAGE 18 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

4. PV22xS Enclosure Overview

The PV22xS SCSI Enclosure provides three configuration topologies, also referred to as
modes, which are end-user configurable: Joined
in Figure 4-1. A 3-position switch located on the Split Bus Module (SBM) sets the desired
enclosure topology. Each of these topologies will be further explained in more detail
below. The configuration will not take effect until the enclosure is power cycled which is
designed to protect from accidental topology changes. SCSI Enclosure Services (SES)
management features are an integral part of the enclosure. SES provides a unique set of
features to monitor critical environmental variables such as temperature and power status.

, Split, and Cluster topologies, as shown

Figure 4-1 PV22xS Topologies

Joined Bus Topology

This is the simplest topology of operation providing access to all 14 hard drives on one
shared SCSI bus. Only one cable must be attached between the server and the primary
Enclosure Management Module (EMM) in the enclosure. If operating with two EMM
modules, which provide redundant SES services, a second cable should not be attached
to the secondary EMM. There can only be one physical path from the enclosure to the
server. Attaching a second cable is not supported and may lead to SCSI bus address
contention. If SES redundancy is not required, a Terminator module can be substituted in
place of secondary EMM.

PAGE 19 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

The primary EMM must always be present. Regardless of whether SES redundancy is
available, the secondary EMM slot must always have either a EMM or Terminator module
installed to properly terminate the SCSI bus. Leaving this slot unpopulated is not
supported and will have undesirable consequences.

Configurations with two EMMs and one cable connected provide a level of redundancy.
Even though there is no data path redundancy, there is enclosure management
redundancy. The redundancy, however, is provided in the form of the SES enclosure
management services to monitor and control critical functions such as temperature, fan
speeds, initiating warnings/alerts, and in the worst case scenarios to safely power down
the enclosure to prevent any potential damage under adverse thermal conditions. SES
functions by default are performed by primary EMM. In the event there is a catastrophic
failure of the primary EMM, the secondary EMM automatically will detect this event and
perform enclosure management functions.

Split Bus Topology

This topology splits the PV22xS into two halves, essentially providing two independent,
electrically isolated, SCSI busses capable of supporting up to seven (7) HDD on each bus.
Because the two busses are electrically isolated, this allows simultaneous SCSI
transactions on each bus. Two cables are required, one connected to each EMM. The
cables can attach to two host systems (Figure 4-2) running different applications and
operate completely independent from one another, or they can be connected to two SCSI
channels on a single host system (Figure 4-3) to increase performance by splitting the I/O
traffic between two different physical data paths In either case, it is highly recommended
that both halves of the PV22xS be connected to the same type of SCSI RAID controllers
(e.g. PERC4 and PERC4 or PERC3 and PERC3). The primary EMM monitors the state
of the seven HDD in slots labeled 0 thru 5, and 8. The secondary EMM monitors the
state of the seven HDD in slots labeled 9 thru 15. SES management is provided by both
EMM modules to their respective SCSI hosts. SCSI failures on one bus will not affect the
other bus. Power and enclosure cooling, however, are shared. Power disruptions (e.g.,
power short on a HDD) will affect both busses equally.

Figure 4-2 Split Bus Cabling to Two Host System

PAGE 20 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

Figure 4-3 Split Bus Cabling Options to a Single Host System, (a) Same Adapter, (b) Different
Adapters

Cluster Topology

Cluster Topology is very similar to Joined bus topology. It provides the capability to have a
single shared SCSI bus with the additional capability of allowing concurrent access from
two host controllers (Figure 4-4). In this topology, only 13 HDD are available since the
PERC controller on the second host system uses up one of the available SCSI addresses
(refer to Figure 4-5 for more details on the specific SCSI ID reassignments).

Dell™ High Availability Clustering

(for more information refer to Appendix – A: References)
provides the capability of configuring the PV22xS Enclosure to attach simultaneously to
two different host systems. The hardware configuration of both host systems, including the
SCSI RAID controller, must be identical. This is ideal for High Availability scenarios where
access to the same data from two different systems is desirable. Clustering falls into two
categories: Active/Active and Active/Passive.

Microsoft® Cluster Service MSCS and all PowerEdge™ Clusters support both
active/active and active/passive cluster configurations. The term active/active refers to a
cluster with at least one virtual server running on each node. For example, in a 2-node
cluster configuration, when an application is running on Node 1, Node 2 does not need to
remain idle waiting for Node 1 to fail. Node 2 can run its own cluster-aware applications (or
another instance of the same application) while providing failover capabilities for resources
on Node 1. Node 1 in turn can provide failover capabilities for resources on Node 2.
Although both nodes have physical access to the same storage, the disk arrays are
mutually exclusive and cannot be simultaneously accessed by both nodes. In the event of

PAGE 21 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

a failover condition, access rights to the disk array assigned to the failed node is
transferred to the other node. An active/active cluster node must be sized appropriately to
handle the load of both nodes (in the event of a failover).

The term active/passive refers to failover cluster configurations in which one or more
cluster nodes is actively processing requests for cluster applications while at least one
cluster node simply waits for an active node to fail. An active/passive configuration is more
costly in terms of price/performance because one or more servers sits idle most of the
time. It is appropriate for business-critical systems since the application can use the full
power of another server in case of a failure.

Dell PowerEdge Clustered systems fully support both cluster topologies with MSCS.

Figure 4-4 Cluster Topology Cabling Example

PV22xS – Common Features and Capabilities

Regardless of what topology is ultimately selected, the PV22xS provides a basic set of
common features and capabilities.

SCSI ID Assignment

SCSI standard requires that all devices on the SCSI bus have a unique address, called a
SCSI ID, with a maximum of 16 devices that can reside on the bus. The SCSI ID for the
enclosure devices are hardwired by the enclosure backplane. PV22xS supports 15 of
these devices with the SCSI controller in the host system using the remaining SCSI ID. In
Cluster topology, however, the enclosure will give up one of its HDD SCSI ID, which will

PAGE 22 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

be reserved for use by the second SCSI controller. A mapping of the SCSI ID address
space reserved by the PV22xS enclosure is shown in Figure 4-5. The HDD identifiers are
directly mapped to it’s assigned SCSI ID, as shown in Figure 4-6.

Figure 4-5 SCSI ID Assignments

Figure 4-6 SCSI ID Numbers and Associated Hard Disk Drives

Multiple PV22xS Enclosures on a Single Host System

Daisy chaining of Multiple PV22xS enclosures is not supported. This is due to SCSI
limitation of only allowing a maximum of 16 devices on one SCSI bus. Daisy chaining
would effectively concatenate the devices in two (or more) enclosures into a single SCSI
bus, thus violating the unique SCSI ID requirement.

Multiple PV22xS enclosures, however, can still be configured for a single host. The
enclosure must be cabled directly to a separate SCSI channel on that host. A SCSI
channel does not have to be on a physically separate adapter. For instance, a mult­channel SCSI controller such as a PERC4/DC is considered to be two SCSI channels
even though both channels reside on one physical PCI adapter. In this example, two
PV22xS can be attached to the PERC4/DC adapter. Conversely, it is also perfectly
acceptable to use two PERC4/DC adapters and attach an enclosure to each. The
maximum number of enclosures that can be configured for a single host system is limited
by the number of channels the host system can support.

Mixing EMM Speeds

There are two types of EMM modules that PV22xS supports: U160 and U320. The
numbers refer to the SCSI bus speed (in MB/sec) that the EMM is capable of supporting.
For example, U160 can transfer up to 160MB/sec while U320 EMM supports up to
320MB/sec transfer rates. One or two EMM modules may be installed in the enclosure.
When two are installed, it is required that they have the same transfer speed, which can
be easily identified with labels on back of EMMs. Mixing of U160 and U320 EMM

PAGE 23 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

modules is not supported. For more details, refer to section 5 (SCSI Host Bus Adapters /
SCSI RAID).

Actual transfer rate of the SCSI bus is determined by host channel, EMMs and HDDs
installed in PV22xS.

Mixing HDD Capacities

Mixing of different drive sizes is supported, but not recommended. PV22xS currently
supports 18GB, 36GB, 73GB, 146GB, and 300GB capacities, with future support for
higher capacities as they become available. If operating in a RAID configuration, the
lowest capacity drive may define the RAID virtual disk size. For example, with two
different HDDs, one 36GB and one 146GB drive, a RAID1 size will be limited to 36GB.
On a RAID0 array size will be limited to 72GB. The excess space that is not included in
the RAID array may still be partitioned and used. For more details on RAID array
configurations, refer to section 5 (SCSI Host Bus Adapters / SCSI RAID).

Mixing HDD Spindle Speeds

Mixing of different drive spindle speeds is supported; however, it is not recommended.
PV22xS currently supports 10K and 15K rpm spindle speeds, with future support for
higher spindle rates as they become available. If operating in a RAID configuration, the
lowest spindle speed may dictate the maximum performance of the RAID group. For
example, in a RAID5 configuration with 5 drives having 4 HDDs running at 15K RPM and
one running at 10K RPM, effective data transfer rate may be limited to the performance
provided by the 10K RPM drive since all HDDs in the array perform as a single entity.

SCSI Enclosure Services (SES)

Environmental parameters such as temperature and voltage are continuously monitored
by the EMM module. The PV22xS works in conjunction with Dell OpenManage™ to
monitor and manage these parameters. Trigger thresholds can be set to warn the user
when these parameters exceed safe operating levels, and to take specific action to power
down the enclosure when they exceed a critical threshold level for maximum safety and
avoid potential damages before they occur. Both local visual indicators and audio alerts
are provided by the enclosure. By default, the audio alerts on the enclosure are disabled
from the factory and must be enabled by the customer using Dell OpenManage, if desired.
In summary, the following SES functions are provided by the PV22xS enclosure:

• Report enclosure shelf faults

• HDD Power/Fault/Offline status

• Power supply status and shutdown capabilities

• Fan speed status and control

• Temperature monitoring

• Over/Under temperature shutdown capabilities

• Alarms (local audible and visual LED indicators)

Warm and Hot Pluggable Features

The PV22xS enclosure is designed for warm and hot plugging to help prevent damage
when pulling or replacing certain components. Warm and hot plugging are very similar
concepts allowing for the safe removal and replacement of components while the power is
still on. The main difference is that warm plugging requires any I/O operations to be

PAGE 24 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

properly quiescent before the remove/replacement operation takes place, whereas hot
plugging can occur at any time without prior notification. Hot plugging allows the user
applications and disk activities to continue uninterrupted while maintenance or repair
actions are taking place. Warm plugging requires synchronization with the software
applications to temporarily halt disk I/O activity before the hardware is removed/replaced.

• Warm Pluggable Devices

o EMM module
o SBM module
o Terminator module

• Hot Pluggable Devices

o HDD (when disk I/O is still occurring, all activities must be halted first)
o Power supply module
o Fan modules

• Neither Warm nor Hot Pluggable Devices
The backplane is critical for proper enclosure operations and cannot be warm or
hot plugged. Power MUST be removed from the system.

External SCSI Cables

Depending on the enclosure topology, one or two external SCSI cables can be attached
between the host system(s) and the PV22xS enclosure. Although cables may physically
look and feel the same, there are actually differences in the cable material and
construction methods used that produces different quality levels of cable. In general, the
higher the SCSI speed, the higher the quality of the cable that is required. SCSI cable
quality is governed by the American National Standards Institute, Inc. (ANSI).

Specifically, for SCSI U160 speeds, a cable must comply with the ANSI SCSI Parallel
Interface-3 (SPI-3) specification. For U320 speeds, ANSI SPI-4 must be adhered to. All
PV22xS SCSI cables purchased from Dell have been tested and comply with the ANSI
SPI-4 standard. When using existing customer cables or if the cables are purchased from

rd

3

party sources, ensure the cables are approved for ANSI SPI-4 operation. Using lower
grade cabling will cause unpredictable results which can compromise the stability and
reliable operation of the PV22xS enclosure.

External SCSI Connector Types

There are two types of SCSI connectors used by the external SCSI cables. One is called
the High Density (HD) connector, also commonly referred to as a P-connector, and the
other is the Very High Density Cable Interconnect (VHDCI). Both are 68-pin connectors,
but the VHDCI has a smaller form factor, as shown in Figure 4-7. Two types of cables are
available for the PV22xS, an HD-to-HD and HD-to-VHDCI. The EMM module uses the HD
style connector where one end of the cable is attached. The other end will have either an
HD or VHDCI connector. PERC3 and PERC4 series controllers use a VHDCI connector.
Older controller cards generally use the HD connector. Always verify the controller
connector type to ensure cables will be compatible.

PAGE 25 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

Cable Connector Controller Connector

P-type connector – Male

VHDCI-type connector – Male

• Figure 4-7 External SCSI Connector

P-type connector – Female

VHDCI-type connector – Female

PV22xS Configuration Topology and Application Attributes

PV22xS provides different configuration topologies, which may impact the storage solution
for an application. All PV22xS configuration modes have various benefits for application
attributes and are summarized as depicted in Table 4-1 and Table 4-2:

• Table 4-1: PV22xS Configuration Topologies Vs Application Attributs

Application

Attributes

Reliability

Joined Split Cluster

Default Default Default

PV22xS Configuration Topologies

Relatively better than
Joined topology –

Availability

Default

depending on RAID

High
level (see Table 4-2 for
details)

Serviceability

All interconnect components are Field Replaceable units

Redundancy

Performance

Data

Protection

Scalability

Cost

• Not on SCSI
Path

• Management
capabilities with
both ZEMMs

• Not on SCSI Path

• Management

capabilities with
both ZEMMs

Provides both data
and management

Relatively lower

Depends on HDD and ZEMM technology –

U160/U320

performance due to

RAID controller

configuration

requirements

Default Default Default

Maximum

Could be lower with

Single ZEMM and

non-redundant PSU

Minimum (for a given
topology)

Requires two ZEMMs, but can be lower with
non-redundant PSU

Same as Joined, but
one HDD less

PAGE 26 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

• Table 4-2: PV22xS Configuration Topologies Vs RAID Level Performance

PV22xS Configuration Topologies

RAID Levels

Joined Split Cluster

RAID 0

Default RAID 0
performance

Best performance, if
configured across multiple
channels

Similar to Joined
topology

Increased Availability, if

RAID 1

Default

mirror set is on separate

Default

channel

RAID 5

Default Default Default

Higher performance than

RAID 10

Default

RAID 1 with same increased

Default
availability
Increase availability, if

RAID 50

Default

spanned across separate

Default
channels

It can be observed that selection of a PV22xS configuration topology impacts not only
Application Attributes but also impacts RAID level performance. Besides these
configuration topologies, PV22xS provides some common capabilities as listed below for
the storage solution:

Redundancy Capabilities

• Physical SCSI Data Path Redundancy

PV22xS using two Enclosure Management Modules (EMM) provides redundant
data path when configured in Cluster topology.

• Management Redundancy

With multiple EMMs in PV22xS, hardware redundancy is provided when servicing
enclosure management requests

• Power Supply Redundancy

PV22xS is capable of running on single power supply; multiple power supplies
provide 1+1 redundancy

• Cooling Redundancy

PV22xS provides support of up to 2 fan units; each fan unit has 2 blowers in all
making 4 fan blowers. The unit is thermally designed to continue operating with 1
of 4 fan blower failure, thus giving a 3+1 cooling redundancy

Reliability Capabilities

• Thermal Reliability

PAGE 27 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

PV22xS consists of four digital temperature probes, which in conjunction with
SES processor provides device protection against adverse thermal conditions.
PV22xS is designed to operate at ambient temperature of 10
the temperature probe detects ambient temperature to be below 0

O

50

C, SES processor notifies host of critical temperature range, which would then
initiate a thermal shut down process. In case, PV22xS temperature probes detect
ambient temperature either to be at or below -5

O

C or above 55OC, SES processor

O

C to 35OC. Once

O

C or above

initiates a self shutdown process and hence protecting the device against adverse
thermal conditions.

Serviceability Capabilities

• All components including backplane easily accessible and replaceable in rack

• LED indicators for visual status reporting on internal components

• Audible alarm to notify critical/non-critical status (for details please refer to

PV22xS user’s guide). Status is also reported to system management tools –
OpenManage – OMSM or Array manager (for details refer to OpenManage user’s
guide)

PAGE 28 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

5. SCSI Host Bus Adapters / SCSI RAID

This section addresses concerns specific to RAID. A summarized table (Table 5-2) is
presented at the end of the section which provides an overview of the Dell SCSI and RAID
solutions that support PV22xS.

Although attachment of PV22xS to 39160 SCSI HBA is supported, it is not recommended.
39160 is a U160 SCSI HBA, which is primarily intended for SCSI Tape devices, and would
limit the performance of SCSI subsystem to U160 speeds. The controller doesn’t provide
any hardware RAID benefits similar to software RAID.

RAID: Hardware Vs Software

A RAID system mostly provides a fail-safe with the capability of improved performance
data storage on a group of hard disk drives. RAID, when implemented with a hardware
engine – including a dedicated RAID processor and memory system – is termed a
hardware RAID. Whereas, when the RAID engine is implemented on a host processor
and memory – either as part of operating system or driver, it is termed a software RAID.

Hardware and Software RAID implementations are associated with their respective pros
and cons. The fact that the Hardware RAID is implemented using dedicated hardware
resources gives a performance boost to the host systems which are required to perform
heavy data processing. Software RAID performance scales based on the available
memory and processing resources for a given host system. Thus, host systems which are
used for heavy data processing will exhibit slower performance with Software RAID when
compared to Hardware RAID implementations.

Selection of Hardware RAID vs Software RAID is a trade-off between performance and
cost. Enterprise host systems which perform heavy data processing and are mission
critical would be more suited for a Hardware RAID implementation. For more details on
Hardware and Software RAID please refer to “Comparing RAID Implementation Methods”
and “Implementing Software RAID on Dell PowerEdge Servers” papers in Appendix – A:
References.

Considerations to Implement a RAID Solution

Implementing a RAID solution requires that certain considerations be addressed in order
to achieve an optimum solution. These considerations are impacted by:

• Controller Family

• Data Protection Strategy

• High Availability Needs - Clustering

• Heterogeneous vs Homogeneous Environment – SCSI bus technology

(U160/U320) and Hard Disk Drive (HDD) spindle speeds

• Interface Channel options - Internal and External channels
o Spanning across channels/controllers/PV22xS

• Capacity Needs – Virtual Disk Size

Each of these above factors is discussed in detail in following sections.

PAGE 29 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

Controller Family

Aside from the technical, physical and feature differences between the controller families
(see Table 5-2), there are a few critical differences which impact behavior of these
controllers and hence a storage solution. Dell’s Storage controllers can be broadly
categorized into two families – Adaptec PERCs and LSI PERCs. Design of these families
impact the customer experience and may lead to ambiguities when configuring more
complex storage strategies/solutions.

Though most of these differences are transparent to the user, when using the
recommended Dell OpenManage™ software to configure storage subsystem; however, in
cases where the BIOS utility is the preferred tool, an in-depth understanding of controller
behavior is required to avoid possible issues. These differences include:

• Controller BIOS utility terminology and operation
o The user interfaces of the boot-time BIOS utilities for the different

controller families differ in layout and terminology. In environments where
controllers from both families are being managed via the BIOS utilities,
these differences can lead to confusion. Refer to user guides for
controller specific information on the functionality and terminology used
by these utilities.

NOTE: In the context of this document, the word ‘array’ (LSI BIOS) =
‘container’ (Adaptec BIOS) = ‘logical drive’ (LSI POST) = ‘volume’
(OpenManage Software).

• Layout of multiple logical arrays on the same group of disks
o LSI family of controllers require that once an array is created on a group

of disks, any subsequent arrays created on any of these same disks can
only be created on the same group of disks; subsequent arrays are not
allowed to use any disks not in the original group, nor any subset of the
original group. Adaptec controllers are less restrictive in this regard.

o Similarly LSI controllers are more restrictive of hot spare functionality in

this scenario. If redundant and non-redundant arrays are placed on the
same group of disks and a hot spare is assigned to the redundant array,
because of the non-redundant array’s failed state, the hot spare will
never start the rebuild on the redundant array.

• Mixed-size hard drives within a RAID10/RAID50
o RAID50 not supported on Adaptec controllers. This configuration is

supported on LSI controllers only

o When mixing drive sizes in RAID10/RAID50 on LSI controllers, with

default maximum size, the controller will not coerce the larger ‘child’
arrays to match the size of the smallest ‘child’ array. Controller will stripe
across all ‘child’ arrays until the smallest one is full, then it will continue to
stripe across the remaining ‘child’ arrays until the next one is full, and so
on until all the available space is used. This can impact performance as
the array fills up, since the number of stripes decreases as ‘child’ arrays
fill up.

PAGE 30 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

In contrast, Adaptec controllers will only use space on all ‘child’ arrays
equal to the available space on the smallest ‘child’ array, and will leave
the remaining space on the larger drives unused and available, and the
performance behavior will not be affected as the array fills up.

• Array Reconfiguration / Disk Expansion / Rebuild
o As stated in the ‘Getting Started with RAID’ guide (see Appendix A –

References), the recommended method of changing an array
configuration is to back up the data on the array, create a new array with
the new desired parameters, and then migrate the data to the new array.
In cases where this is not possible, reconfiguring arrays is supported for
the controllers. However, should an array need to be reconfigured, the
reconfiguration paths supported differ on different controllers. Please
refer to the documentation on each controller for information on
supported reconfiguration paths.

o Many factors may influence the time it takes to perform a rebuild or

reconfiguration operation on an array including: Size of disk and array,
RAID level, IO profile, system load, PERC family,

For more details on the rebuild process please refer to Appendix A –
Reference 5: Maintenance Best Practices for Direct Attached SCSI
Solutions

• Controller upgrade / Array Migration / Drive Roaming / metadata
o PERC controllers store array configuration metadata both in controller

NVRAM and on-disk. Arrays are upgradeable and can be migrated within
the family of controllers made by the same vendor; i.e., migration of
arrays from one LSI controller to another or from one Adaptec controller
to another is possible whereas migration between the two families is not;
e.g., an array from an LSI controller cannot be migrated to an Adaptec
controller, or vice-versa.

o Drive roaming is initiated when the HDDs are changed to different SCSI

IDs or channels on a same controller. When the HDDs are placed on
different SCSI IDs or different channels of the same controller, controller
detects the RAID configuration from configuration metadata residing on
the HDD. Drive roaming is supported on all families of SCSI RAID
controllers.

NOTE: In a clustering environment, drive roaming is supported within the
same channel only.

o For additional information on Array Migration or Drive Roaming, please

refer to the User Guide for the controller.

• Write cache policy considerations
o Write back

In write-back caching, data transfer is completed when the controller
cache receives all data from host for the write transaction. Write-back
caching is faster, but on PERC controllers that do not have a battery

PAGE 31 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

back up unit (see Table 5-2), there is a risk of data loss if system power
is lost while unwritten information is still in the cache.

o Write Through

In write-through caching, a data transfer is completed when the disk
subsystem receives all of the data from the host. There is no risk of data
loss in a power-loss scenario (as in write back), but it is slower, since the
controller must wait for the hard drive to return a good status on the write
to the controller before proceeding to the next operation.

If the PERC controller has a battery backup unit (see Table 5-2) or if a
host system is powered by an alternate power source (e.g. UPS etc), the
controller’s cache retains data during a power loss. Thus PERCs with
battery back-up not only provide the performance benefit of write-back
caching but also help ensure integrity of data during system power loss.
For data retention time please refer to Table 5-2.

Data Protection Strategy

• Battery back up: see “Write Cache policy considerations” section above.

• Global hot spare

o A global hot spare can be used for any array that is on the same

controller as the hot spare.

o More cost-effective than dedicated hot spares, but does not protect

multiple arrays as completely as dedicated hot spares.

o A global hot spare is selected randomly and may be in a different

enclosure than the failed drive it is replacing. Can impact performance if
rebuilt array spans enclosures/channels.

• Dedicated hot spare

o A dedicated hot spare is assigned to one or more arrays.

o Dedicated hot spares that reside in the same storage enclosure typically

have better performance than global hot spares.

o Dell recommends using dedicated hot spares for critical data.

• Consistency checks

o Consistency checks verify the correctness of redundant data in an array.

o As during a rebuild, array performance will be degraded while check is

running.

o Some PERC controllers allow pausing of consistency check and resume

it later or to resume the consistency check after the system reboots.

o Recommended as part of regular system maintenance for critical data.

PAGE 32 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

For more details on rebuild process please refer to Appendix A –
Reference 5: Maintenance Best Practices for Direct Attached SCSI
Solutions

• Patrol Read
o Patrol Read is a preventative maintenance background operation

(available only on PERC 3 (except PERC 3/DI), PERC 4 and PERC 4e
controller families running 3.0 and higher firmware).

o Patrol Read examines each block in a configured physical disk for media

errors and attempts to fix media errors by reallocating the bad block.

o Patrol Read is designed to run so as to not impact I/O performance.

o Patrol Read has two run modes: Manual and Automatic. The mode is set

from within the controller BIOS. By default the run mode is set for
“Automatic”.

o Highly recommended as part of regularly scheduled system

maintenance to prevent downtime due to media errors.

NOTE: Patrol Read and Consistency Check are not the same operations;
Patrol Read examines the physical disk for media defects while Consistency
Check validates data using data parity, to help prevent data errors.

High Availability Needs – Clustering

• Clustering is supported only on some HBAs – see Table 5-2. Clustering is not

supported on embedded or non-RAID controllers.

• Multiple logical volumes on the same physical array are not supported in

clustering environment – only one logical volume/array.

Heterogeneous vs. Homogeneous environments

• SCSI Technology (U320/U160/Ultra 2)

It is possible to mix a combination of hard drives compliant with U320, U160
or Ultra-2 generation of SCSI technologies. Mixing of hard drives is supported
as long as all the devices on the SCSI bus have Low Voltage Differential
(LVD) signaling.

Though possible, it is not recommended to mix drives, since overall SCSI bus
performance will be limited by the slowest technology present in the mix of
array. Where mixing of HDD technology is not avoidable, it is a best practice
to have all hard drives in any particular array be of matching technology. In all
cases, the overall subsystem performance will be limited to some degree by
the slowest technology present in the mix.

• Spindle Speed

It is recommended that all hard drives within any particular array be of the
same spindle speed (RPM – which can vary from 10,000-15,000 on the hard

PAGE 33 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

drives supported by the PV22xS). Mixing spindle speeds within the same
array will impact array performance, especially in a striped (RAID0, RAID10
or RAID50) environment. Higher RPM drives will have to wait for the slower
ones to complete the operations. This is the case regardless of cache policy
settings.

Interface Channel Options – Internal/External/Multi channel device configuration

• See Table 5-2 for internal/external connector details for each controller.

• Do not connect devices both internally and externally on any one channel.

• Termination

o The PV22xS is also already terminated. No external SCSI bus

terminators are needed when connecting to a PV22xS.

• Spanning of Arrays

Dell’s PERC controllers provide additional array configuration options of
spanning logical arrays. Dell supports spanning of logical arrays only across
multiple channels on same PERC controllers. If multiple PV22xS are
configured (Joined or Split topology - Figure 4-3a) on the same PERC then a
logical array may be spanned across multiple PV22xS.

In general, spanning of logical arrays across multiple channels on a controller
improves data throughput. With the introduction of another data path,
depending on the RAID level, the configuration may compromise data
availability. When configuring logical arrays across multiple channels it is
essential to understand the trade-off between performance, reliability and
availability of data. Depending on the RAID level for an array, spanning may
offer a better solution than without spanning. The following table lists pros
and cons of spanning in relation to RAID levels:

PAGE 34 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

• Table 5-1: Pros and Cons for Spanning Arrays with various RAID Levels

Level Pros Cons

RAID 0

RAID 1

Improves data throughput by
optimizing bandwidth utilization

Improves data availability
against channel failures, if
mirrored set is spanned across
separate channel

Reduces reliability and availability of data
with additional point of failure – second
channel
If PV22xS is used on both channels
(either in Split of Joined topology), limits
the usage of PERC on internal channel
(connecting devices to the same channel
on internal and external connectors is
not supported)

RAID 5

RAID 10

RAID 50

May improve data throughput for
certain accesses, if data and
parity for the access span
across channels

Improves data availability
against channel failures, if
mirrored set is spanned across
separate channel

May improve data throughput,
similar to spanned RAID5

Even though RAID5 provides data
protection but compromises reliability
and availability of data at SCSI channel
level
If PV22xS is used on both channels
(either in Split of Joined topology), limits
the usage of PERC on internal channel
(connecting devices to the same channel
on internal and external connectors is
not supported)
Reduces reliability and availability of data
as compared to non-spanned RAID50 at
SCSI channel level

Note: Dell PERC controllers don’t support spanning of arrays across different
controllers. Spanning across channels on the same PERC controller is only
supported.

Capacity Needs – Virtual Disk Size

Virtual disks on SCSI PERC controllers cannot be created with an aggregate
size greater than 2TB due to limitations on SCSI controllers and operating
systems. Host management application (OMSM, Array Manager) will give an
error if an attempt is made to make virtual disk greater than 2TB, prompting
user to reduce size of virtual disk. For more information please refer to
“Beyond The 2-TB SCSI Logical Unit” white paper at:

http://www.dell.com/downloads/global/vectors/2004_2tblun.pdf

Additional RAID information

o Dell ‘Getting Started with RAID’ document

http://support.dell.com/support/edocs/storage/RAID/RAIDbk0.pdf

.

PAGE 35 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

Table 5-2 – Controller Families & Features: Part 1

PERC 4e/Di PERC 4e/DC

SCSI
Technology

Host Bus
Interface

# Channels

# internal
connectors

# External
Connectors

RAID Levels
Supported

Cluster
Support?

Cache Size
(DIMM)

Write Cache
Options

Read Cache
Options

BBU Battery Life
Max # Logical

Drives/Controller

Wide
Ultra320

x4 PCI
Express
(Embedded)

2 2 2 2

2 x 68-pin
HD

1 x 68-pin
UHD

0, 1, 5, 10, 50 0, 1, 5, 10, 50 0, 1, 5, 10, 50 0, 1, 5, 10, 50

no yes no no

256MB
DDR2

off, write­through,
write-back
read-ahead
(default),
off
(selectable)

48-72 hours 48-72 hours 48-72 hours 48-72 hours

40 40 40 40

Wide
Ultra320

X8 PCI
Express
(HBA)

2 x 68-pin
HD

2 x 68-pin
UHD

128MB DDR

off, write­through,
write-back
read-ahead
(default),
off
(selectable)

Max # Physical
Drives/Logical

32 32 32 32

Drive
Vendor
SCSI Chip

LSI LSI LSI LSI

LSI 53C1030 LSI 53C1030 LSI 53C1030 LSI 53C1030

PERC 4/Di

(PE2600)

Wide
Ultra320

PCI 2.2
(Embedded)

2 x 68-pin
HD

1 x 68-pin
HD

128MB
DIMM

off, write­through,
write-back
read-ahead
(default),
off
(selectable)

PERC 4/Di

(PE1750)

Wide
Ultra320

32-bit / 33
MHz PCI
(Embedded)

n/a
(hardwired to
backplane)

1 x 68-pin
UHD

128MB
DDR2

off, write­through,
write-back
read-ahead
(default),
off
(selectable)

RAID Processor

Intel GC80332 Intel GC80332 Intel GC80303 Intel i80321

PAGE 36 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

Table 5-2 – Controller Families and Features: Part 2

PERC 4/SC PERC 4/DC PERC 3/SC PERC 3/DCL

SCSI
Technology

Host Bus
Interface

# Channels

# internal
connectors

# External
Connectors

RAID Levels
Supported

Cluster
Support?

Cache Size
(DIMM)

Write Cache
Options

Read Cache
Options

BBU Battery Life
Max # Logical

Drives/Controller
Max # Physical

Drives/Logical
Drive

Vendor
SCSI Chip

Wide Ultra320 Wide Ultra320

64-bit / 66 MHz
PCI 2.2 (HBA)

64-bit / 66 MHz
PCI 2.2 (HBA)

Wide Ultra3 /
U160

64-bit / 66 MHz
PCI 2.2 (HBA)

Wide Ultra3 /
U160

64-bit / 66 MHz
PCI 2.2 (HBA)

1 2 1 2

2 x 68-pin HD 2 x 68-pin HD 1 x 68-pin HD 2 x 68-pin HD

1 x 68-pin UHD 2 x 68-pin UHD 1 x 68-pin UHD 2 x 68-pin UHD

0, 1, 5, 10, 50 0, 1, 5, 10, 50 0, 1, 5, 10, 50 0, 1, 5, 10, 50

no yes no No

64 MB 128MB 64 or 128MB 64 or 128MB

off, write­through,
write-back
read-ahead
(default), off
(selectable)

off, write­through,
write-back
read-ahead
(default), off
(selectable)

off, write­through,
write-back
read-ahead
(default), off
(selectable)

off, write­through,
write-back
read-ahead
(default), off
(selectable)

no BBU 48-72 hours no BBU no BBU

40 40 40 40

32 32 32 32

LSI LSI LSI (AMI) LSI (AMI)

LSI 53C1020 LSI 53C1030

Qlogic 12160 Qlogic 12160

RAID Processor

Intel GC80302 Intel GC80303

Intel i960 Intel i960

PAGE 37 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

Table 5-2 – Controller Families and Features: Part 3

PERC 3/DC PERC 3/QC PERC 3/Di PERC 2/QC

SCSI Technology

Host Bus Interface

# Channels

# internal
connectors

# External
Connectors

RAID Levels
Supported

Cluster Support?

Cache Size

Write Cache
Options

Read Cache
Options

BBU Battery Life
Max # Logical

Drives/Controller
Max # Physical

Drives/Logical
Drive

Vendor

SCSI Chip

RAID Processor

Adaptec

39160

Wide Ultra3 /
U160

64-bit / 66
MHz PCI 2.2
(HBA)

Wide Ultra3 /
U160

64-bit / 66
MHz PCI 2.2
(HBA)

Wide Ultra3 /
U160

64-bit / 66
MHz PCI 2.2
(Embedded)

Wide Ultra 2

64-bit / 33
MHz PCI 2.1
(HBA)

Wide
Ultra3 /
U160
64-bit / 66
MHz PCI

2.2 (HBA)

2 4 2 4 2

2 x 68-pin
HD
(LVD); 1x
50-pin
Ultra

2 x 68-pin
HD

2 x 68-pin
HD

2 x 68-pin HD
(except 1650
& 2650 ­N/A)

2 x 68-pin
HD

0, 1 or 2 x 68-

2 x 68-pin
UHD

4 x 68-pin
UHD

pin HD
(server
platform

4 x 68-pin
UHD

2 x 68-pin
UHD
(LVD)

dependant)

0, 1, 5, 10, 50 0, 1, 5, 10, 50 0, 1, 5, 10 0, 1, 5, 10

n/a

yes no no no no

64 or 128MB
DIMM

off, write­through,
write-back
read-ahead
(default),
off
(selectable)

64 or 128MB
DIMM

off, write­through,
write-back
read-ahead
(default),
off
(selectable)

128MB DIMM

off, write­through,
write-back
read-ahead
(default),
off
(selectable)

16 - 128 MB
DIMM

off, write­through,
write-back
read-ahead
(default),
off
(selectable)

n/a

n/a

n/a

48-72 hours 48-72 hours 48-72 hours 48-72 hours n/a

40 40 64 64 n/a

32 32 64 64 n/a

LSI (AMI) LSI (AMI) Adaptec Adaptec Adaptec

Qlogic 12160

2 X Qlogic
12160

Intel i960 Intel i960

Adaptec 7899

Intel i960 OR
GC80303
(server
platform
dependant)

2 x Adaptec
7897N

StrongARM
SA110S

Adaptec
7899

n/a (non­RAID)

PAGE 38 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

6. Summary and Conclusions

After analyzing and assessing the importance of all the storage components relative to
attributes essential for performance of any given application, it can be observed that not
only understanding the characteristics of each storage component is critical but also how
the storage component is configured has direct impact on the effectiveness of an
application.

Based on the study of each storage component relative to attributes essential for
effectiveness of an application, Table 6-1 lists typical storage recommendations for various
applications. Recommendations listed in the table are only an example for a storage
solution; each actual storage solution will have additional factors which need to be
considered before determining an optimum storage solution. These additional factors
include:

• Data and power back-up strategies of an Enterprise IT

• Storage deployment environment:

o Number of active and passive users
o Type of data – mission critical or non-mission critical

• Storage deployment budget

PAGE 39 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

• Table 6-1: Typical Storage Recommendation for Different Storage Applications

Application

Servers Key Parameters RAID Level PV22xS Host Controller

* Availability

Email

Database

Web * Performance

On-line
Transaction

File * Scalability

Streaming
Media

* Performance
* Scalability

* Availability
* Performance

* High
Availability

* Performance

RAID10
Provides both
Availability and
performance

RAID10
Provides highest
availability of all
other RAID levels

RAID0 in Web
Server farms
Provides maximum
performance of all
RAID levels

RAID10
Provides highest
availability of all
other RAID levels

RAID5
Maximizing storage
with some data
protection

RAID0
Provides maximum
performance of all
RAID levels

Storage Recommendations

For added
Cluster Topology
Scalable and
high availability

Cluster Topology
Provides
availability of all
PV22xS
topologies

Split Topology
Spanning RAID0
across channels
provides best
performance

Cluster Topology
Provides
availability of all
PV22xS
topologies

Joined Topology
Provides
maximum
scalability

Split Topology
Spanning RAID0
across channels
provides best
performance

performance

enable Write back

cache, if available

on PERC

If added

performance is

required enable

Write back cache,

if available on

PERC

PERC, with read

ahead capabilities

for added

performance boost

For added

performance

enable Write back

cache, if available

on PERC and data

protection at

PERC level

PERC with Global

Hot Spare will

improve

availability for

RAID5

PERC, with read

ahead capabilities

for added

performance boost

Remarks

Recommendation targets a
typical Enterprise Email
Server

Performance requirements
for DB servers depend on
number of users

Web server farms, usually
have multiple servers
providing replicated web
data. Thus, data availability
is handled at server level

Recommendation targets
Enterprise applications
which require higher
availability than
performance.
In general for all OLT data
protection is of utmost
importance
These servers usually have
data backed up periodically
and hence, require minimum
data protection from Storage
components
Performance is of utmost
importance for Streaming
media servers. Data on
these servers is usually
backed up on external
medias such as Tape
libraries or optical medias ­CD/DVD

PAGE 40 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

7. Appendix – A: References

1. ANSI T10/1365D SCSI Parallel Interface – 4 (SPI-4) and SCSI Block Commands – 2

(SBC-2)

(http://www.t10.org

2. Dell™ PowerVault™ 220S and 221S Systems Documents

)

(http://support.dell.com/support/edocs/stor-sys/spv22xs/en/index.htm

3. Dell ‘Getting Started with RAID’ document

(http://support.dell.com/support/edocs/storage/RAID/RAIDbk0.pdf

4. Dell High Availability Cluster

(http://www1.us.dell.com/content/topics/global.aspx/solutions/en/clustering_h

a?c=us&cs=555&l=en&s=biz)

5. Maintenance Best Practices for Direct-Attached SCSI Solutions

(http://www.dell.com/downloads/global/products/pvaul/en/Maintenance%20B

est%20Practices%20for%20Direct-Attached%20SCSI%20Solutions.pdf)

6. Beyond The 2-TB SCSI Logical Unit

(http://www.dell.com/downloads/global/vectors/2004_2tblun.pdf

7. Comparing RAID Implementation Methods

)

)

)

(http://www1.us.dell.com/content/topics/global.aspx/power/en/ps3q03_dumouc

helle?c=us&cs=555&l=en&s=biz)

8. Implementing Software RAID on Dell PowerEdge Servers

(http://www1.us.dell.com/content/topics/global.aspx/power/en/ps3q03_nadon?c

=us&cs=555&l=en&s=biz)

PAGE 41 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

8. Appendix – B: Glossary

Array – Collection of multiple physical Hard Disk Drives to form a single logical volume

BBU – Battery Backup Unit

BIOS – Basic Input Output System

Cluster – Multi-initiator Environment with MSCS operating environment

HBA – Host Bus Adapter (non-RAID controllers)

HDD – Hard Disk Drive

Hot Plug – Insertion or removal of device without the need to quiescent I/Os

PERC – Power Edge RAID Controller

RAID – Redundant Array of Inexpensive Disks or Redundant Array of Independent Disks

ROMB – RAID On Motherboard

SCSI – Small Computer System Interface

U160 – Ultra 160 (160MB/sec throughput)

U320 – Ultra 320 (320MB/sec through put)

Warm Plug – Insertion or removal of device with all the I/Os quiescent

PAGE 42 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

9. Appendix – C: About the Dell SCSI Storage Solution Team

Authors

David Papasan – International Product Support
Gary Mott – PowerVault™ 22xS Product Engineering Manager
John Woo – PowerVault 22xS Product Engineering Lead
Mickey McGee – RAID Engineer
John Langdon – RAID Engineer
Phil Curran – PowerVault 22xS Marketing
Ryan King – International Product Support
Samit Ashdhir – PowerVault 22xS Product Engineer

PAGE 43 11/17/2005

A REFERENCE GUIDE FOR OPTIMIZING DELL™ SCSI SOLUTIONS VER A02

10. Appendix – D: Revision History

Revision Description Date Modified By

A00 First Release 04/28/2005

A01

Updated “™” and “®”
information where applicable

05/06/2005 Samit Ashdhir

SCSI Storage

Solution Team

Updated requirement for
controllers to be attached to
PV22xS in Split and Cluster

A02

Topologies (Section 4)
Updated with brief on “Hardware
Vs Software RAID” (Section 5)

11/17/2005

SCSI Storage

Solution Team

Updated “Drive Roaming” and
“Patrol Read” (Section 5)

PAGE 44 11/17/2005