HSJ80 Array Controller
ACS Version 8.5J-2
Configuration Planning Guide
Second Edition (August 2000)
Part Number: EKHSJCPPA. B01
Compaq Computer Corporation
2000 Compaq Computer Corporation.
COMPAQ , the Compaq logo, and StorageWorks Registered in the U.S. Patent and Trademark Office.
NonStop and OpenVMS are trademarks of Compaq Information Technologies Group, L.P.
Microsoft, Windows, and Windows NT are registered trademarks of Microsoft Corporation.
Confidential computer software. Valid license from Compaq required for possession, use, or copying.
Consistent with FAR 12.211 and 12.212, Commercial Computer Software, Computer Software
Documentation, and Technical Data for Commercial Items are licensed to the U.S. Government under
vendor's standard commercial license.
Com paq shall not be liable for technical or editorial errors or omissions contained herein. The
information in this document is subject to change without notice.
THE INFORMATION IN THIS PUBLICATION IS PROVIDED "AS IS" WITHOUT WARRANTY OF
ANY KIND. THE ENTIRE RISK ARISING OUT OF THE USE OF THIS INFORMATION REMAINS
WITH RECIPIENT. IN NO EVENT SHALL COMPAQ BE LIABLE FOR ANY DIRECT,
CONSEQUENTIAL, INCIDENTAL, SPECIAL, PUNITIVE OR OTHER DAMAGES WHATSOEVER
(INCLUDING WHITHOUT LIMITATION, DAMAGES FOR LOSS OF BUSINESS PROFITS,
BUSINESS INTERRUPTION OR LOSS OF BUSINESS INFORMATION), EVEN IF COMPAQ HAS
BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES AND WHETHER IN AN ACTION
OF CONTRACT OR TORT, INCLUDING NEGLIGENCE.
The limited warranties for Compaq products are exclusively set forth in the documentation
accompanying such products. Nothing herein should be construed as constituting a further or
additional warranty.
Printed in the U.S.A.
Compaq StorageWorks HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
Second Edition (August 2000)
Part Number EKHSJCPPA. B01
Contents
About This Guide
Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x
Text Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x
Special Notices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
Related Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii
Revision History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii
Chapter 1
Planning a Subsystem
Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Controller Designations A and B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Controller Designations: this and other . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Failover . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Configuration Rules. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Selecting a Cache Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Read Caching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Read-Ahead Caching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Write-Through Caching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Write-Back Caching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Non-Volatile Memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Fault-Tolerance for Write-Back Caching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Mirrored Caching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Dynamic Caching Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Assigning Unit Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Cluster Interconnect Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
iv HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
Single or Dual Controller, Single Cluster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Dual Controller, Dual Host Port, Single Cluster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Dual Controller, Multiple-Cluster Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Chapter 2
Planning Storage
Where to Start . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Device PTL Addressing Convention. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Determining Storage Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Choosing a Container Type. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Creating a Storageset Profile. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Storageset Planning Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Stripeset Planning Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Mirrorset Planning Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
RAID set Planning Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
Striped Mirrorset Planning Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
Partition Planning Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
Defining a Partition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
Guidelines for Partitioning Storagesets and Disk Drives. . . . . . . . . . . . . . . . . . . . . . . . . . . 220
Loader Planning Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Choosing Switches for
Storagesets and Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Enabling Switches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Changing Switches. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
Storage Switch Types. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
RAID set Switches. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
Mirrorset Switches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
Partition Switches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
Device Switches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
Initializing Switches. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
Chunk Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
Save Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
Destroy or Nodestroy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
Other Switches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
Partitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
Maximum Cache Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
Read Cache . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
Run . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Write Protection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Write-Back Cache. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Storage Maps. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Contents v
Creating a Storage Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
Example Storage Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
Using the LOCATE Command to Find Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
The Next Step . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
Appendix A
Subsystem Profiles
Storageset Profile. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A2
Type of Storageset:. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A2
Enclosure Templates for Device Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A3
Glossary
Index
vi HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
Figures
Figure 11. Controller Locations and Terminology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Figure 12. Mirrored Caching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Figure 13. Single Controller, Single Host Port, Single Cluster Configuration . . . . . . . . . . . . . 19
Figure 14. Dual Controller, Single Host Port, Single Cluster Configuration. . . . . . . . . . . . . . 110
Figure 15. Dual Controller, Dual Host Port, Dual Star Coupler Configuration. . . . . . . . . . . . 111
Figure 16. Dual Controller, Dual Host Port, Multiple-Cluster Configuration . . . . . . . . . . . . . 112
Figure 21. PTL Naming Convention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Figure 22. Example PTL Addressing with a Dual-Redundant Controller Configuration . . . . . 24
Figure 23. Mapping a Unit to Physical Disk Drives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Figure 24. Container Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Figure 25. Sample Storageset Profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Figure 26. A Three-Member RAID 0 Stripeset (Example 1). . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Figure 27. A Three-Member RAID 0 Stripeset (Example 2). . . . . . . . . . . . . . . . . . . . . . . . . . 210
Figure 28. Members Distributed Across Controller I/O Device Ports. . . . . . . . . . . . . . . . . . . 212
Figure 29. Mirrorsets Maintain Two Copies of the Same Data . . . . . . . . . . . . . . . . . . . . . . . . 213
Figure 210. Mirrorset Example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
Figure 211. First Mirrorset Members Placed on Different SCSI Port Addresses . . . . . . . . . . 214
Figure 212. Three-Member RAIDset Using Parity (Example 1) . . . . . . . . . . . . . . . . . . . . . . . 215
Figure 213. Five-Member RAIDset Using Parity (Example 2) . . . . . . . . . . . . . . . . . . . . . . . . 216
Figure 214. Striped Mirrorset (Example 1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
Figure 215. Striped Mirrorset (Example 2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
Figure 216. Example of a Partitioned Single Disk Unit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
Figure 217. Chunk Size Larger Than the Request Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
Figure 218. Chunk Size Smaller Than the Request Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
Figure 219. Sample Storage Map. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Tables vii
Tables
Table 21 A Comparison of Container Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Table 22 Example Chunk Sizes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
About This Guide
This document presents the configuration planning guidelines for the HSJ80 Array
Controller and storagesets running Array Controller Software (ACS) Version 8.5J-2.
The first chapter describes the configuration rules within which the HSJ80 Array
Controller needs to operate. It also provides a description of the available cabling options
to the cluster. The selection of a particular cabling option affects how you will configure
the controller to operate within the cluster.
The second chapter contains information to guide you in the selection of the types of
storage for the subsystem. Together, these two chapters should enable you to plan the total
subsystem configuration needed for your business.
This book does not contain information about the operating environments to which the
controller may be connected, nor does it contain detailed information about subsystem
enclosures or their components. See the documentation that accompanied these hosts and
peripherals for their respective information.
x HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
Conventions
This book uses the text conventions and special notices described in the following
sections.
Text Conventions
Convention Meaning
UPPERCASE Command syntax that must be entered exactly as shown, for example:
SET FAILOVER COPY=OTHER_CONTROLLER
UPPERCASE CLI command name discussed within text, for example:
"Use the SHOW SPARESET command to show the contents of the
spareset."
Monospaced Screen display as shown in the text.
sans serif italic Command variable or numeric value that you supply, for example:
SHOW RAIDset-name or
or
serif italic SET THIS_CONTROLLER CHUNKSIZE=n
serif italic Reference to other book titles, for example:
See the HSJ80 Array Controller Troubleshooting Resources Guide for
details.
. Indicates that a portion of an example or figure has been omitted.
About This Guide xi
Special Notices
This book does not contain detailed descriptions of standard safety procedures. However,
it does contain warnings for procedures that could cause personal injury, and cautions for
procedures that could damage the controller or its related components. Look for these
symbols when you are carrying out the procedures in this book:
WARNING: A Warning contains information essential to human safety. It advises
users that failure to take or avoid a specific action could result in physical harm to the
user. Use a warning, not a caution, when such damage is possible.
CAUTION: A Caution contains information that the user needs to know to avoid
damaging the software or hardware.
IMPORTANT: An Important note is a type of note that provides information essential to the
completion of a task. Users can disregard information in an important note and still complete a
task, but some error or failure may result.
NOTE: A Note emphasizes or supplements important points of the main text. It supplies
information that may apply only in special cases--for example, memory limitations, equipment
configurations, or details that apply to specific versions of a program.
xii HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
Related Publications
The following table lists some of the documents related to the use of the controller, cache
module, and external cache battery.
Document Title 6-3 Part Number 2-5-2 Part Number
Compaq StorageWorks HSJ80 Array Controller ACS N/A AA-RN17A-TE
Version 8.5J-2 Installation and Configuration Guide
Compaq StorageWorks HSJ80 Array Controller ACS N/A EK-H SJSO-OA
Version 8.5J-2 Solution Software Kit Overview
Compaq StorageWorks HSJ80 Array Controller ACS N/A EK-H SJAA-RA
Version 8.5J-2 Release Notes
Compaq StorageWorks HSJ80 Array Controller ACS N/A EK-H SJCL-RA
Version 8.5J-2 CLI Reference Guide
Compaq StorageWorks HSJ80 Array Controller ACS N/A EK-J80MS-SA
Version 8.5J-2 Maintenance and Service Guide
Compaq StorageWorks HSJ80 Array Controller ACS N/A EK-J80TR-SA
Version 8.5J-2 Troubleshooting Resources Guide
Compaq StorageWorks HSJ80 Array Controller ACS N/A EK-H SJAN-AA
Version 8.5J-2 Command Console 2.3 Application Notes
Revision History
First Edition: December 1999
Second Edition: August 2000
1
Chapter
Planning a Subsystem
This chapter contains concepts that will help you plan your subsystem:
s "Terminology" on page 11
"Controller Designations A and B" on page 12
"Controller Designations: this and other" on page 12
s "Failover" on page 13
s "Configuration Rules" on page 13
s "Selecting a Cache Mode" on page 14
s "Assigning Unit Numbers" on page 18
s "Cluster Interconnect Options" on page 18
When you have conceptually planned the subsystem, you can use Chapter 2 as a guide to
plan the subsystem storage devices. You will then need to use the plans developed from
Chapters 1 and 2 to configure both the controller and subsystem storage. The
configuration procedures to use are located in the Compaq StorageWorks HSJ80 Array
Controller ACS Version 8.5J-2 Installation and Configuration Guide.
Terminology
When configuring the subsystem, you will encounter the following terms and concepts
that you must understand:
s controller A and controller B
s this controller and other controller
12 HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
Controller Designations A and B
Controllers and cache modules are designated either A or B, depending on their location
within the storage enclosure. The relationship is a physical one and is illustrated in
Figure 11, which shows a horizontal mounting of the controller enclosure. Vertical
mountings have the controller pair on the visual left: A = left, B = right.
Controller Designations: this and other
Some CLI commands use the terms this and other to identify one controller or the other in
a dual-redundant pair. These designations are defined as follows:
s this controller--the controller that is the focus of the CLI session. That is, the
controller through which the CLI commands are being entered (may be controller A or
B). The maintenance (local) terminal is connected to the maintenance port of this
controller.
s other controller--the controller that is not the focus of the CLI session and through
which CLI commands are not being entered. The maintenance terminal is not
connected to the other controller.
The relationship is a relative one defined by the location of the maintenance port cable, as
illustrated in Figure 11.
1
2
3
CXO6997A
1 Controller A (other controller in this sample maintenance port connection)
2 Controller B (this controller in this sample maintenance port connection)
3 Controller Maintenance Port (with serial cable attached)
Figure 11. Controller Locations and Terminology
Planning a Subsystem 1 3
Failover
Dual-redundant controller configurations have a feature called failover. Failover keeps the
storage array available to the cluster in the event of a single-controller failure by allowing
the surviving controller to take control of the entire subsystem.
CI bus-based subsystems have both controllers connected to the same storage units.
Because both controllers service the same storage units, either controller can continue to
service subsystem storage if its companion controller fails.
If you configure the subsystem storage devices before setting the controllers to a failover
condition, make sure you know which controller has the good storage configuration
information before entering the CLI command SET FAILOVER COPY=this_controller.
This command places the configuration information into the "other" controller.
Configuration Rules
Before you can configure controller storage, you will need to review the following
configuration rules to ensure your configuration meets the requirements and conditions:
s Maximum 36 physical storage devices using DS-BA356-S and DS-BA356-MW
StorageWorks enclosures (dual-redundant controller configurations)
s Maximum 42 physical storage devices using DS-BA356-S and DS-BA356-MW
StorageWorks enclosures (single-controller configurations)
s Maximum 72 storage devices in the Model 4200 enclosure
s Maximum 2 host ports per HSJ80 controller
s Maximum 16 nodes per cluster
s Maximum 32 nodes per cluster when using a Computer Interconnect Star Coupler
Extender (CISCE) on the cluster
s Maximum 512 GB unit capacity
s Maximum 8 partitions per storageset or individual disk
s Storageset Maximums (general)
24 members per storageset
30 RAID-1 storagesets
20 RAID-5 storagesets
45 RAID-0/1/5 storagesets
30 RAID-1/5 storagesets
14 HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
20 RAID-5 and RAID-1 storagesets for single controller configurations
Up to 6 members per mirrorset (RAID-1)
From 2 to 24 members per stripeset (RAID-0)
From 3 to 14 members per RAIDset (RAID-3/5)
48 devices per mirrored stripeset (24 devices x 2)
Selecting a Cache Mode
Before selecting a cache mode, you should understand the caching techniques supported
by the cache module. The cache module supports the following caching techniques to
increase the performance of the subsystem read and write operations:
s Read caching
s Read-ahead caching
s Write-through caching
s Write-back caching
These cache modes are described in the following paragraphs.
Read Caching
Read caching decreases the subsystem response time to a read request by allowing the
controller to satisfy the read request from the cache memory, rather than from the disk
drives.
When the controller receives a read request from the host, it reads the data from the disk
drives, delivers it to the host, and stores it in cache memory. If the host requests the same
data again, the controller can satisfy the read request from the cached data, rather than
re-reading it from the disk drives. By default, read caching is enabled for all storage units.
Refer to the commands SET unit MAXIMUM_CACHED_TRANSFER and SET unit
MAX_READ_CACHED_TRANSFER_SIZE in the Compaq StorageWorks HSJ80 Array
Controller ACS Version 8.5J-2 CLI Reference Guide for more detail.
Read-Ahead Caching
Read-ahead caching begins once the controller has processed a read request and it receives
a sequential read request from the host. If the controller does not find the data in the cache
memory, it reads the data from the disks and sends it to the cache memory.
Planning a Subsystem 1 5
The controller then anticipates subsequent read requests and fetches the next block or
blocks of data from the disks as it sends the requested read data to the host. This is a
parallel action. The controller notifies the host of the read completion; subsequent
sequential read requests are satisfied through the cache memory. By default, read-ahead
caching is enabled for all disk units. Refer to the command SET unit
READAHEAD_CACHE in the Compaq StorageWorks HSJ80 Array Controller ACS
Version 8.5J-2 CLI Reference Guide for more detail.
Write-Through Caching
Write-through caching also decreases the subsystem response time to a read request by
allowing the controller to satisfy the request from the cache memory rather than from the
disk drives.
When the controller receives a write request from the host, it stores the data in its cache
memory, writes the data to the disk drives, then notifies the host when the write operation
is complete. This process is called write-through caching because the data actually passes
through--and is stored in--the cache memory on its way to the disk drives. If the host
requests the recently written data, the controller satisfies the read request from its cache
memory, rather than from the disk drives.
If read caching is enabled for a storage unit, write-through caching is also enabled. Also,
because both caching techniques enhance the controller's read performance, write-through
caching is automatically disabled when you disable read caching.
By default, read caching (and, therefore, write-through caching) is enabled for all storage
units.
Write-Back Caching
This caching technique decreases the subsystem response time to write requests by
allowing the controller to declare the write operation complete as soon as the data reaches
its cache memory. The controller performs the slower operation of writing the data to the
disk drives at a later time.
By default, write-back caching is disabled for all storagesets. The controller will not
provide write-back caching to a unit unless you ensure that the cache memory is
non-volatile as described in the "Non-Volatile Memory" section that follows.
Refer to the commands SET unit MAXIMUM_CACHED_TRANSFER and SET unit
MAX_WRITE_CACHED_TRANSFER_SIZE in the Compaq StorageWorks HSJ80
Array Controller ACS Version 8.5J-2 CLI Reference Guide for more detail.
16 HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
Non-Volatile Memory
The controller cannot provide write-back caching to a unit unless its cache memory is
non-volatile. In other words, you must provide a back-up power source to the cache
module to preserve the unwritten cache data in the event of a power failure. If the cache
memory is volatile--that is, if it has no back-up power--the unwritten cache data is lost
during a power failure.
By default, the controller expects to use an external cache battery (ECB) as the cache
module back-up power source to enable it to be non-volatile (see the Compaq
StorageWorks HSJ80 Array Controller ACS Version 8.5J-2 Maintenance and Service
Guide for more information about the ECB).
If the subsystem is backed up with a UPS as an alternative to the ECB, a switch must be
set in the CLI command SET this_controller UPS=DATACENTER_WIDE or SET
this_controller UPS=NODE_ONLY or SET this_controller NOUPS. See the Compaq
StorageWorks HSJ80 Array Controller ACS Version 8.5J-2 CLI Reference Guide for
details about using the CLI Command: SET this_controller UPS.
NOTE: The controller executes multiple write operations to satisfy a single write request for a
RAIDset or mirrorset. For this reason, a RAIDset or mirrorset requires non-volatile cache
memory to ensure data redundancy until the write request is satisfied.
Regardless of the back-up power source you choose, the cache-memory power LED
flashes about once every three seconds to indicate the cache module memory array is
receiving power from its primary power source.
Fault-Tolerance for Write-Back Caching
The cache module supports the following features to protect the availability of its
unwritten (write-back) data:
s Non-volatile memory (required for write-back caching)
s Mirrored caching (optional)
s Dynamic caching techniques (automatic)
Mirrored Caching
To further ensure the availability of unwritten cache data, you can use a portion of each
cache module memory to mirror the other cache module's write-back data in a
dual-redundant configuration. Refer to the command SET this_controller (or
other_controller) MIRRORED_CACHE command in the Compaq StorageWorks HSJ80
Array Controller ACS Version 8.5J-2 CLI Reference Guide for more detail.
Planning a Subsystem 1 7
Figure 12 shows the principle of mirrored caching: half of cache A mirrors cache B
write-back data and vice versa. This arrangement ensures that the write-back data is
preserved if a cache module or any of its components fail.
NOTE: When your controllers are configured to use mirrored write-back cache, the cache
capacity is half of the total amount of cache in the configuration. If each cache module has 512
MB of cache for a total of 1024 MB of cache in the configuration, the cache capacity is 512 MB.
Cache module A Cache module B
A B
cache cache
Copy of Copy of
B A
cache cache
CXO5729A
Figure 12. Mirrored Caching
Before configuring dual-redundant controllers and enabling mirrored write-back cache,
make sure the following conditions are met:
s Both controllers have the same size cache (512 MB).
s Diagnostics indicate that both cache modules are good.
s Both cache modules have a battery present (unless you have enabled the UPS switch).
A battery does not have to be present for either cache if you enable the UPS switch.
s No unit errors are outstanding; for example, lost data or data that cannot be written to
devices.
s Both controllers are started and configured in failover mode.
For important considerations when replacing or upgrading memory in a mirrored cache
configuration, see the Compaq StorageWorks HSJ80 Array Controller ACS Version 8.5J-2
Maintenance and Service Guide.
18 HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
Dynamic Caching Techniques
If the controller detects a full or partial failure of its cache module or ECB, it
automatically reacts to preserve the cached write-back data. Then, depending upon the
severity of the failure, the controller chooses an interim caching technique--also called
the cache policy--which it uses until you repair or replace the cache module or ECB.
Assigning Unit Numbers
Every container (storageset, partition, or JBOD disk) needs a unit number to communicate
to the hosts. Each unit number contains the following:
s A letter that indicates the device type in the storage unit.
D for disk drives (including optical drives)
T for tape drives
s A number that must be in the range 0 through 4094, which uniquely identifies the
storage unit (4095 and 4096 are reserved).
The unit number is placed into the system with the ADD UNIT unit-number
container-name command. This command assigns a unit number to the container name
established with the ADD DISK or ADD TAPE command.
For a detailed description of these commands, refer to the Compaq StorageWorks HSJ80
Array Controller ACS Version 8.5J-2 CLI Reference Guide.
Cluster Interconnect Options
The HSJ80 may be cabled to the cluster in several more options than were previously
available with HSJ40 and HSJ50 subsystems. The supported cabling options are listed
below and described in the paragraphs that follow:
s Single HSJ80 Array Controller, single host port, single Star Coupler, single cluster
s Dual HSJ80 Array Controllers, single host port, single Star Coupler, single cluster
s Dual HSJ80 Array Controllers, dual host port, dual Star Couplers, single cluster
s Dual HSJ80 Array Controllers, dual host port, dual Star Couplers, multiple clusters
Planning a Subsystem 1 9
Single or Dual Controller, Single Cluster
A single or dual HSJ80 controller subsystem configured to use only one of the two
available host ports functions in a manner similar to that of the earlier HSJ40 and HSJ50
subsystems. This mode of operation requires that the second port (the unused port) on the
controller be disabled.
Either CI port may be used, but the port not being used must be disabled using the
SET this_controller NOPORT_1_PATH_A (or NOPORT_2_PATH_B) CLI command.
Figure 13 shows the single controller configuration and Figure 14 the dual-controller
configuration.
This cabling configuration is the basic HSJ80 upgrade option; there are no subsystem
configuration changes required.
HSJ80
CI port 1 or 2
Star Coupler
Adapter Adapter
Host Host
CXO7065A
Figure 13. Single Controller, Single Host Port, Single Cluster Configuration
110 HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
HSJ80 HSJ80
CI port 1 or 2 CI port 1 or 2
Star Coupler
Adapter Adapter
Host Host
CXO7066A
Figure 14. Dual Controller, Single Host Port, Single Cluster Configuration
Dual Controller, Dual Host Port, Single Cluster
This cabling configuration may be used to correct the problem in which the host data rate
exceeds the capability of a single host adapter. Adding the second host adapter and cabling
it to the second HSJ80 host port (by way of a second Star Coupler) allows a parallel path
to storage.
The dual controller, dual host port, dual Star Coupler, dual host adapters option allows a
single cluster to effectively double the throughput of storage available to it. This option
also improves availability by being capable of surviving a failure of either an HSJ80 CI
host port or host adapter. Figure 15 shows the configuration.
Planning a Subsystem 1 11
HSJ80 HSJ80
CI port 2 CI port 1 CI port 2 CI port 1
Star Coupler
Adapter Adapter Adapter Adapter
Host Host
CXO7067A
Figure 15. Dual Controller, Dual Host Port, Dual Star Coupler Configuration
Dual Controller, Multiple-Cluster Configuration
The approved cabling method for a multiple-cluster subsystem configuration is shown in
Figure 16. This configuration gives the subsystem better throughput, better space
utilization, and reduced maintenance cost. The reduced maintenance cost is derived from
the use of a single dual-redundant HSJ80 subsystem being capable of replacing two
HSJ50 storage subsystems.
CAUTION: When using this mode of operation, the port access of the controllers and
the logical units must be set correctly or data loss may occur. See the following
description for additional information.
An HSJ80 Array Controller storage subsystem may be used in a multiple-cluster
configuration using dual Star Couplers and both host ports on both controllers cabled in a
dual-redundant configuration.
112 HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
HSJ80 HSJ80
CI port 2 CI port 1 CI port 2 CI port 1
Star Coupler
Adapter Adapter
Host Host
Adapter Adapter
Host Host
Cluster B Cluster A
CXO7068A
Figure 16. Dual Controller, Dual Host Port, Multiple-Cluster Configuration
Multiple-cluster configurations should be configured with caution because data stored on
a disk with one cluster may be read and written by the other cluster if not configured
properly. To avoid the problems that this could cause, you must:
s Select the containers to be used with a particular cluster.
s Enter a series of CLI commands for each container in the subsystem to make them
accessible by only one of the two clusters.
The SET unit_number THIS_PORT_1_ACCESS= FULL (or NONE) CLI command
assigns specific storage units for access by only one of the two clusters. Each container in
the subsystem must be set up in the manner described in the next section.
Planning a Subsystem 1 13
The following example of a set of commands shows the setting of the access to port 1 to
full on this controller (and other) for container D100. This is followed by the syntax to
turn on the access to port 2 to full on this controller (and other) for container D200:
Set D100 THIS_PORT_1_ACCESS=FULL
Set D100 THIS_PORT_2_ACCESS=NONE
Set D100 OTHER_PORT_1_ACCESS=FULL
Set D100 OTHER_PORT_2_ACCESS=NONE
Set D200 THIS_PORT_2_ACCESS=FULL
Set D200 THIS_PORT_1_ACCESS=NONE
Set D200 OTHER_PORT_2_ACCESS=FULL
Set D200 OTHER_PORT_1_ACCESS=NONE
Using Figure 16 as the example, cluster A now has full access to D100, while cluster B
does not have access to D100. Conversely, cluster B now has full access to D200, while
cluster A does not have access to D200.
2
Chapter
Planning Storage
This chapter provides information to help you plan the storage configuration of your
subsystem. Use the guidelines found in this section to plan the various types of storage
containers needed.
The following information is contained in this chapter:
s "Where to Start" on page 22
s "Device PTL Addressing Convention" on page 23
s "Determining Storage Requirements" on page 25
s "Choosing a Container Type" on page 26
s "Creating a Storageset Profile" on page 27
s "Storageset Planning Considerations" on page 29
"Stripeset Planning Considerations" on page 29
"Mirrorset Planning Considerations" on page 212
"RAID set Planning Considerations" on page 215
"Striped Mirrorset Planning Considerations" on page 217
s "Partition Planning Considerations" on page 219
s "Loader Planning Considerations" on page 221
s "Choosing Switches for Storagesets and Devices" on page 221
s "Storage Maps" on page 229
s "The Next Step" on page 232
22 HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
Where to Start
Containers are defined as individual disk drives (JBODs), storagesets of varying types
(mirrorsets, stripesets, and so on), and partitioned drives. The following is a structure you
can follow to plan your storage configuration. The references in each step locate details
about specific commands and concepts. Appendix A provides blank templates that you
may use to keep track of the containers being configured.
1. Familiarize yourself with the current physical layout of the devices and their
addressing scheme. See "Device PTL Addressing Convention" on page 23.
2. Determine your storage requirements. Use the questions in "Determining Storage
Requirements" on page 25 to help you.
3. Choose the type of storage containers you need in your subsystem. See "Choosing a
Container Type" on page 26 for a comparison and description of each type of
storageset.
4. Create a storageset profile (described in "Creating a Storageset Profile" on page 27).
Fill out the storageset profile while you read the sections that pertain to your chosen
storage type:
s "Storageset Planning Considerations" on page 29
"Stripeset Planning Considerations" on page 29
"Mirrorset Planning Considerations" on page 212
"RAID set Planning Considerations" on page 215
"Striped Mirrorset Planning Considerations" on page 217
s "Partition Planning Considerations" on page 219
5. Decide which CLI command switches should be turned on for your subsystem. CLI
command device switches apply to all devices, including those configured as
single-disk units (JBOD). General information on switches is detailed in "Choosing
Switches for Storagesets and Devices" on page 221. Detailed information is also
provided in the Compaq StorageWorks HSJ80 Array Controller ACS Version 8.5J-2
CLI Reference Guide.
s Determine which unit switches you want for your units (see "Device Switches" on
page 223).
s Determine which initialization switches you want for your planned storage
containers (see "Initializing Switches" on page 223).
6. Create a storage map (see "Storage Maps" on page 229).
Planning Storage 2 3
7. Configure the storage you have now planned using one of the following methods:
s StorageWorks Command Console (SWCC) graphical user interface (GUI).
s Command Line Interpreter (CLI) commands by way of a terminal or PC connected
to the maintenance port of the controller. This method allows you greater flexibility
in defining and naming storage containers. The Compaq StorageWorks HSJ80
Array Controller ACS Version 8.5J-2 CLI Reference Guide provides CLI command
details.
Device PTL Addressing Convention
The controller has six I/O ports (SCSI device ports), each of which connects to a SCSI
bus. Each SCSI bus connects to a storage enclosure that supports up to seven storage
devices (targets) in the BA35x enclosure or up to 14 in the 4200 series enclosure. In
dual-controller subsystems, these device buses are shared between the two controllers.
NOTE: The SWCC graphical user interface (GUI) calls the device ports "channels."
The controller identifies storage devices based on the SCSI Port-Target-LUN (PTL)
numbering scheme. The physical location of a storage device in its enclosure determines
its PTL:
s P--Designates the controller's I/O port number (1 through 6).
s T--Designates the target identification (ID) number of the device. Valid target ID
numbers are 0 through 15.
s L--Designates the logical unit number (LUN) of the device. For disk devices, the
LUN is always 0. Passthrough devices may have LUNs other than 0.
PTL addressing is used in the CLI commands to add device types (Disk, CD-ROM, and so
on).
Figure 21 shows the structure of a PTL address.
1 02 00
LUN 00
Target 02 (range = 0-15
Port 1 (range = 1-6
Figure 21. PTL Naming Convention
24 HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
Figure 22 shows a sample SA500/SA800 rack arrangement of four device enclosures,
each cabled to an I/O cable from the controller enclosure (I/O ports 1 through 4). Note that
the connection of the I/O cables are in sequence from top to bottom on the illustration (the
suggested manner). The target addresses are fixed within the device enclosures, with 0 on
the right and 6 on the left (looking at the front of the SA500/SA800). Detailed information
about this example is located in "Creating a Storage Map" on page 230.
Target
5 4 3 2 1 0
1
PORT
Po we r Power D 106 D 105 D 104 D103 D 101 D 100
Supply Supply RAID2 S1 S1 RAID1
M1
M3
Disk10500 Disk10400 Disk10300 Disk10200 Disk 10100 Disk10000
2
D 101 D 100
D103
D104
Po we r Power D 107 D 105
RAID1
S1
S1
Supply Supply RAID2
M3 M1
Disk20200
Disk20300
Disk20400 Disk20000
Disk20100
Disk20500
3
D 100
Po we r Power D 108 D 105 D104 D103 D 102
RAID1
Supply Supply RAID2 S1 S1
M2
M4
Disk30100 Disk30000
Disk30500 Disk30400 Disk30300 Disk30200
4
Po we r Power D 109 D 105 D104 D103 D 102 Spare
Supply Supply RAID2 S1 S1
M4 M2
Disk40500 Disk40400 Disk40300 Disk40200 Disk40100 Disk40000
Figure 22. Example PTL Addressing with a Dual-Redundant Controller Configuration
Planning Storage 2 5
When the controller receives an I/O request, it identifies the storageset unit number for the
request, and reconciles the unit number to the storageset name. From the storageset name,
the controller locates the appropriate devices for the I/O request. For example, the storage
unit named D100 (Figure 23) is a RAIDset named "RAID1." RAID1 in this example
contains DISK10000, DISK20000, and DISK30000. The controller generates the read or
write request to the appropriate devices using the PTL addressing convention.
D100 Host addressable
unit number
RAID1 Storageset
name
Controller
Disk 10000 Disk 20000 Disk 30000
PTL addresses
CXO6186B
Figure 23. Mapping a Unit to Physical Disk Drives
Determining Storage Requirements
You cannot adequately plan your subsystem storage without determining what your
storage requirements are. Here are a few questions you should ask about subsystem usage:
s What applications or user groups will access the subsystem? How much capacity do
they need?
s What are the I/O requirements? If an application is data-transfer intensive, what is the
required transfer rate? If it is I/O-request intensive, what is the required response time?
What is the read/write ratio for a typical request?
s Are most I/O requests directed to a small percentage of the disk drives? Do you want
to keep it that way or balance the I/O load?
26 HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
s Do you store mission-critical data? Is availability the highest priority or would
standard backup procedures suffice?
Choosing a Container Type
Different applications may have different storage requirements, so you will probably want
to configure more than one kind of container within your subsystem.
In choosing a container, you choose between independent disks (JBODs) or one of several
storageset types. The independent disks and the selected storageset may also be
partitioned (see Figure 24).
Storagesets implement RAID (Redundant Array of Independent Disks) technology.
Consequently, they all share one important feature: each storageset, whether it contains
two disk drives or ten, looks like one large virtual disk drive to the host.
Containers
Partition
Stripeset Striped mirrorset RAIDset
Single Mirrorset
(R0) (R0+1) (R3/5)
devices (R1)
(JBOD)
Storagesets
CXO6677A
Figure 24. Container Types
Planning Storage 2 7
Table 21 compares the different kinds of containers to help you determine which ones
satisfy your requirements.
Table 21 A Comparison of Container Types
Container Name Relative Request Rate Transfer Rate Applications
(Read/Write) (Read/Write) MB
Availability
per second
I/O per second
Independent disk Equal to number of Identical to single disk Identical to single disk
drives (JBOD) JBOD disk drives drive drive
Stripeset Proportionate to Excellent if used with Excellent if used with High performance for
number of disk drives; large chunk size small chunk size non-critical data
(RAID 0)
worse than single disk
drive
Mirrorset Excellent Good/Fair Good/Fair System drives; critical
files
(RAID1)
High request rates,
RAIDset Excellent Excellent/Poor Read: excellent (if used
read-intensive, data
with small chunk
(RAID 3/5)
lookup
sizes)
Write: good (if used
with small chunk
sizes)
Striped Mirrorset Excellent Excellent if used with Excellent if used with Any critical
large chunk size small chunk size response-time
(RAID 0+1)
application
For a comprehensive discussion of RAID, refer to The RAIDBOOK--A Source Book for
Disk Array Technology (published by the RAID Advisory Board; ISBN # 1-879936-90-9).
Creating a Storageset Profile
Creating a profile for your storagesets, partitions, and devices can help simplify the
configuration process. Filling out a storageset profile helps you choose the storagesets that
best suit your needs and make informed decisions about the switches that can be enabled
for each storageset or storage device configured in your subsystem.
See the sample storageset profile shown in Figure 25.
Appendix A contains blank profiles that you can copy and use to record the details for
your storagesets. Use the information in this chapter to help you make decisions when
creating storageset profiles.
28 HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
Type of Storageset:
_____ Mirrorset __X_ RAIDset _____ Stripeset _____ Striped Mirrorset ____ JBOD
Storageset Name R1
Disk Drives D10300, D20300, D30300, D40300, D50300, D60300
Unit NumberD101
Partitions:
Un it # Un it # Un it # Un it # Un it # Un it # Un it # Un it #
% % % % % % % %
RAIDset Switches:
Reconstruction Policy Reduced Membership Replacement Policy
_X_Normal (default) _X _No (default) _X_Best performance (default)
___Fast ___Yes, missing: ___Best fit
___None
Mirrorset Switches:
Replacement Policy Copy Policy Read Source
___Best performance (default) ___Norm al (default) ___Least busy (default)
___Best fit ___Fast ___Round robin
___None ___Disk drive:
Initialize Switches:
Chunk size Save Configuration Metadata
_X_ Automatic (default) ___No (default) _X_Destroy (default)
___ 64 blocks _X_Yes ___Retain
___ 128 blocks
___ 256 blocks
___ Other:
Unit Switches:
Read Cache Write Cache Maximum Cache Transfer
_X_ Yes (default) ___Yes (default) _X_32 blocks (default)
___ No _X_No ___Other:
Write-Back Cache Write Protection Availability
_X_No (default) _X_Run (default)
___Yes (default)
Yes ___NoRun
_X_No
Host Access Enabled: __________________________________________
Figure 25. Sample Storageset Profile
Planning Storage 2 9
Storageset Planning Considerations
This section provides guidelines to help you choose storageset types for your subsystem:
s Stripeset Planning Considerations
s Mirrorset Planning Considerations
s RAID set Planning Considerations
s Striped Mirrorset Planning Considerations
Stripeset Planning Considerations
Stripesets (RAID 0) enhance I/O performance by spreading the data across multiple disk
drives. Each I/O request is divided into small segments called chunks. These chunks are
then simultaneously "striped" across the disk drives in the storageset, thereby allowing
several disk drives to participate in one I/O request.
For example, in a three-member stripeset that contains disk drives Disk10000, Disk20000,
and Disk30000, the first chunk of an I/O request is written to Disk10000, the second to
Disk20000, the third to Disk30000, the fourth to Disk10000, and so forth until all of the
data has been written to the drives (see Figure 26).
6
1
5
2 4
3
Disk 10000 Disk 20000 Disk 30000
4
Chunk 1 2 3 6
5
CXO5507A
Figure 26. A Three-Member RAID 0 Stripeset (Example 1)
210 HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
The relationship between the chunk size and the average request size determines whether
striping maximizes the request rate or the data-transfer rate. You can set the chunk size or
use the default setting (see "Chunk Size" on page 223 for information about setting the
chunk size). Figure 27 shows another example of a three-member RAID 0 stripeset.
A major benefit of striping is that it balances the I/O load across all of the disk drives in
the storageset. This can increase the subsystem performance by eliminating the hot spots
(high localities of reference) that occur when frequently accessed data becomes
concentrated on a single disk drive.
Virtual disk
Block 0
Operating Block 1
system Block 2
view
Block 3
Block 4
Block 5
etc.
Disk 1 Disk 2 Disk 3
Actual Block 0 Block 1 Block 2
device Block 3 Block 4 Block 5
mappings etc. etc. etc.
Stripeset
CXO4592B
Figure 27. A Three-Member RAID 0 Stripeset (Example 2)
Keep the following in mind as you plan your stripesets:
s Reporting methods and size limitations prevent certain operating systems from
working with large stripesets. The Compaq StorageWorks HSJ80 Array Controller
ACS Version 8.5J-2 Release Notes and Chapter 1 in this Guide ("Configuration Rules,"
page 13) both contain details about these restrictions.
Planning Storage 2 11
s A storageset must contain only disk drives of the same capacity. The controller limits
the capacity of each member to the capacity of the smallest member in the storageset
(base member size) when the storageset is initialized. Thus, if you combine 2-GB disk
drives with 1-GB disk drives in the same storageset, you will waste 1-GB of capacity
on each 2-GB member.
If you need high performance and high availability, consider using a RAIDset,
striped-mirrorset, or a host-based shadow of a stripeset.
s Striping does not protect against data loss. In fact, because the failure of one member
is equivalent to the failure of the entire stripeset, the likelihood of losing data is higher
for a stripeset than for a single disk drive.
For example, if the mean time between failures (MTBF) for a single disk is one hour,
then the MTBF for a stripeset that comprises N such disks is l/N hours. As another
example, if the MTBF of a single disk is 150,000 hours (about 17 years), a stripeset
comprising four of these disks would have an MTBF of only slightly more than four
years.
For this reason, you should avoid using a stripeset to store critical data. Stripesets are
more suitable for storing data that can be reproduced easily or whose loss does not
prevent the system from supporting its critical mission.
s Evenly distribute the members across the device ports to balance load and provide
multiple paths as shown in Figure 28.
s Stripesets for Array Controllers may contain between two and 24 members.
s Stripesets are well-suited for the following applications:
Storing program image libraries or run-time libraries for rapid loading.
Storing large tables or other structures of read-only data for rapid application
access.
Collecting data from external sources at very high data transfer rates.
s Stripesets are not well-suited for the following applications:
A storage solution for data that cannot be easily reproduced or for data that must be
available for system operation.
Applications that make requests for small amounts of sequentially located data.
Applications that make synchronous random requests for small amounts of data.
212 HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
Device ports
1
2
3
54
6
Bus 1
10
2
3
4
5
6
Bus 2
Bus 3
1
5
0
0
1
2
3
4
5
6
2
5
0
0
1
2
3
4
5
6
3
5
0
CXO7073A
Figure 28. Members Distributed Across Controller I/O Device Ports
By spreading the traffic evenly across the SCSI buses (I/O device ports), you ensure that
no one bus (port) handles the majority of data to the storageset.
Mirrorset Planning Considerations
Mirrorsets (RAID 1) use redundancy to ensure availability, as illustrated in Figure 29.
For each primary disk drive, there is at least one mirror disk drive. Thus, if a primary disk
drive fails, its mirror drive immediately provides an exact copy of the data. Figure 210
shows a second example of a Mirrorset.
Planning Storage 2 13
Disk 10000
Disk 10100
A A'
Disk 20000
Disk 20100
Mirror drives contain
copy of data
B'
B
Disk 30100 Disk 30000
C C'
CXO5511A
Figure 29. Mirrorsets Maintain Two Copies of the Same Data
Virtual disk
Block 0
Operating
system Block 1
view Block 2
etc.
Disk 1 Disk 2
Actual Block 0 Block 0
device Block 1 Block 1
mappings Block 2 Block 2
etc. etc.
Mirrorset
CXO4594B
Figure 210. Mirrorset Example 2
214 HSJ80 Array Controller ACS Version 8.5J-2 Configuration Planning Guide
Keep these points in mind as you plan your mirrorsets:
s Data availability with a mirrorset is excellent but comes with a high cost--you need
twice as many disk drives to satisfy a given capacity requirement. If availability is
your top priority, consider using dual-redundant controllers and redundant power
supplies.
s You can configure up to 12 mirrorsets per controller or pair of dual-redundant
controllers. Each mirrorset may contain up to six members.
s Both write-back cache modules must be the same size.
s If you are using more than one mirrorset in your subsystem, you should put the first
member of each mirrorset on different buses, as shown in Figure 211. The first
member of a mirrorset is the first disk drive you add.
When a controller receives a request to read or write data to a mirrorset, it typically
accesses the first member of the mirrorset. If you have several mirrorsets in your
subsystem and their first members are on the same bus, that bus will be forced to
handle the majority of traffic to your mirrorsets.
2
1
1 First member of first mirrorset
| AA-RN17A-TE
DS-BA356-MW
EK-J80MS-SA
EK-J80TR-SA
XO-5135A-MC
XO-5172A-MC
|
