HP D7171A - NetServer - LPr User manual

1
+3ý1HW6HUYHUý $FFHVVRULHV
ZKLWHýSDSHU
ZKLWHýSDSHUZKLWHýSDSHU ZKLWHýSDSHU
PD\ýìäää
6HUYHUý&RQWUROOHUVýDQGý0DVVý6WRUDJHý'HYLFHV
Executive Summary
Description
The proper selection of controller and storage devices, as part of either a
single server or clustered system, enhances the system’s performance and
availability. Storage capacity, data transfer rates and response times to user
requests depend upon the types of controller and storage devices. These
devices significantly impact traditional measurements of system performance.
Availability depends upon the architecture of storage within a server system
so that the data and applications can be both protected and available at a
level consistent with the criticality of the server environment.
Storage management concerns the tools used to monitor and predict the
storage environment. The proper tools can make storage management
intuitive and informative, ultimately reducing total cost of ownership. These
management tools, combined with the correct controllers and disks, can
accommodate eventual system growth in both user base and application
complexity.
Topics Covered
You should read this white paper to become better acquainted with the
important issues related to selecting controller and mass storage devices
including
•Choices between various devices, such as controllers, tape backup
drives, and disk drives
•Physical means of storage organization, i.e. internal, external,
rack-mountable
•Storage topology to improve system availability and performance
•Storage management
This paper reviews many of the storage solutions that are available for
HP NetServers including hard disk drive products, Small Computer System
Interface (SCSI) and disk array controllers, tape backup solutions, and the
range of supported RAID solutions. The paper also examines the newest
offerings including I2O, Fibre Channel (FC), and Ultra2 SCSI where HP has
provided industry-wide leadership in the development and acceptance of
these standards.

ZKLWHýSDSHU
2
+3ý1HW6HUYHUý$FFHVVRULHV
By the time you’ve read this white paper, many of your questions about storage and
controllers will have been addressed. Then the decisions that you make and the
configurations that you select, as you evaluate server system choices, will meet your needs
now and in the future. If you have additional questions on related topics, there are many
HP NetServer white papers and technology briefs that may be of interest to you as well. A
detailed listing of these references and their URLs are included at the end of this paper.
A Closer Look
Purpose
Mass storage is the lifeblood of a server system. Selection of storage components impacts
both performance and availability of the server system (see Figure 1). Storage provides the
capacity to keep programs, applications, data, and other files that the system and the users
require. Storage is also one of the ways to
improve and extend a server system’s
availability. Depending upon the system’s
needs for access, a variety of storage
solutions provide redundancy for the data
and failover to alternative means of
running applications should a failure occur
within a server system. Other aspects of
storage also affect the performance of a
server system. Because mass storage is
the part of a server architecture entrusted
with safeguarding important data and
applications, selection of the appropriate
mass storage components is a primary
issue.
Types of Storage Devices
There are several categories of mass storage devices that are available for configuration in
a server system. Hard disk drives come in a variety of capacities and rotation speeds and
can be used internally or externally to the CPU system. They can be either fixed or hot
swappable. Hot swappable drives, unlike fixed drives, can be removed and replaced
without shutting the system down, thus improving overall system availability. Tape backup
drives are segmented into three basic technologies: Quarter-Inch Cartridges (QIC)/Travan,
Digital Data Storage (DDS) or Digital Audio Tape (DAT) and Digital Linear Tape (DLT). The
performance of the disk drives depends upon the seek time, platter rotation speed, track
density, and size of the disk cache buffer. The performance of the tape drives is a function
of the type of device, device capacity, and data transfer rate. The capacity of both types of
devices enhances the function of the server system. Small server systems currently require
mass storage devices in the 9 GB range while large enterprise systems can require a
terabyte or more of storage in a large arrayed set of drives.
Form/Function Types
Capacity
Access speeds
Internal or External
Availability
Improvements
Performance
Improvements
Back-up
Hot Swap & Hot Spare
Array, RAID, Duplexing
Fibre Channel
Increased Revolutions per Minute
(RPMs)
SCSI
Fast and Wide, Wide Ultra 2
I2O
Figure 1. Analysis of Aspects of Storage

ZKLWHýSDSHU
3
+3ý1HW6HUYHUý$FFHVVRULHV
Connectivity
A SCSI interface is an independent and intelligent local I/O bus. Through this bus a variety
of different devices and one or more controllers can communicate and exchange
information. The SCSI interface operates independently of the actions of the rest of the
system. Many types of peripherals can be
connected along the bus including tape
drives, hard disk drives, scanners,
printers, disk array subsystems, and
CD-ROM drives. Device performance is a
function of the interface data transfer
rates and the saturation of the I/O bus.
Various levels of performance using a
SCSI controller and I/O bus are available
(see Figure 2).
As SCSI technology has evolved, the
server systems benefit by the increased
performance and capability that the most
recent designs provide. The latest step in
the evolution of SCSI is Wide Ultra2 SCSI
which uses low-voltage differential
technology to extend the current SCSI specifications for cable length, transfer speed and
maximum number of devices per bus. The continued evolution of the SCSI standard and its
on-going implementation in products, make it a cost-effective storage solution, especially
for the entry and mid-range level servers.
Fibre Channel (FC), another connectivity standard, is making inroads into the high-end
Intel-based server market as an alternative to SCSI in configurations in which high
availability, storage scalability and configuration flexibility are demanded. Fibre Channel
provides the user with several improvements over the current connectivity technologies.
The server systems can be cabled over much larger distances, eventually up to
10 kilometers, allowing for mirrored configurations that enhance disaster tolerant scenarios.
The server systems will be more scaleable. SCSI supports a maximum of 15 disks per
channel while up to 126 disks can be attached to a single Fibre Channel arbitrated loop.
Fibre Channel cabling is much easier to use than SCSI cabling since terminators are not
required and thin cables with simple connectors replace thick cables with multi-pin
connectors. Another characteristic that distinguishes Fibre Channel from SCSI is that it is
designed to support a network infrastructure, including hubs and switches. Added to
cabling distance, the networking characteristic gives Fibre Channel tremendous flexibility to
support not only high availability system configurations but also storage networking
applications.
0
10
20
30
40
50
60
70
80
90
Ultra SCSI Wide Ultra SCSI Wide Ultra2 SCSI (LVD)
MBytes/Sec
Figure 2. Storage Controller Maximum
Transfer Rate

ZKLWHýSDSHU
4
+3ý1HW6HUYHUý$FFHVVRULHV
Controllers
Controllers (or adapters) are devices that manage the communication interface by adapting
or bridging the data from one I/O bus to another. As an added benefit, controllers can also
organize and protect the data stored on the disks through mirroring, duplexing and/or RAID
techniques. The implementation of these data protection techniques is discussed in the
next section of this paper and on page 8.
Controllers are available in three different form factors:
•Embedded on a motherboard
•As a card in a standard bus slot
•Incorporated into an external storage unit
Examples of each of these forms include:
•Embedded
−SCSI I/O chips
−Embedded RAID in the HP NetServer LH 3 and LH 4 using SCSI I/O chips and a
microprocessor
•Adapter Cards
−HP NetRAID which adapts (or bridges) between PCI and SCSI and also provides
RAID
−HP FC HBA (Fibre Channel Host Bus Adapter) which adapts between PCI and FC
−HP SCSI HBA which adapts between PCI and SCSI
•Incorporated in an external storage unit
−HP FC Array which adapts between FC and SCSI and also provides RAID

ZKLWHýSDSHU
5
+3ý1HW6HUYHUý$FFHVVRULHV
Internal versus External Storage
Internal storage means that the drives are located within the server’s chassis. External
storage allows the storage to be located in a “storage box” that is separate from the CPU.
Servers come equipped to support a certain level of internal storage, usually selected as
functions of user needs and costs. For example, a high-end server system might support a
dozen internal hot-swap disk drives whereas an entry-level system might have only a
couple of fixed internal drives. External storage can be rack-mounted or as a pedestal.
Some high-end systems use only rack-mounted storage for greatest flexibility.
When internal storage is provided, it is typically offered with built-in SCSI controllers,
channels, and backplanes. The SCSI logic is implemented directly on the server system
board. In some configurations, hardware is
available for either mirroring or duplexing
(see Figure 3). Both terms refer to methods
of data protection. Mirroring duplicates the
stored data simultaneously on two separate
disks. If one disk fails, the other remains
available, minimizing data loss and system
interruptions. Duplexing duplicates not only
the data on a disk but also the disk controller,
protecting the data and increasing system
availability in case of either a disk or
controller failure.
External storage is most often preferred for
server failover configurations. When storage devices are externally connected to the
server, they are connected through SCSI or FC cabling, using an external Input/Output
(I/O) slot. The cabling can be either fiber-optic or copper. External storage can also be
used to expand storage capacity for systems with limited internal storage. For example, in a
two-node Microsoft Cluster server configuration, all storage that is shared between the two
system nodes must be external to the server. When considering a rack-mounted storage
system, you should look for the greatest density of storage modules within the standard
EIA rack sizing since floor space is a valuable commodity. Modularity is also important
since rack space may also be used for hubs, routers, and other accessories in any
combination.
Availability and Storage
Disk drives can be organized to provide high availability and data protection, as part of the
data storage function. If the drives are organized as JBOD (just a bunch of disks), then the
system has the storage capacity of those disks but no operational alternatives if one of the
disks goes down and no protection for the data.
Alternatively, any one of several RAID (Redundant Array of Independent Disks) techniques
allow sets of drives to be logically grouped in ways which reduce the drive space overhead,
while still providing comprehensive data protection. RAID systems can be implemented
through either hardware or software.
Single Channel Dual Controllers
1 controller, 1 channel 2 single channel controllers
Figure 3. SCSI Channel Configurations
Mirroring Duplexing
D=data M=mirror
DM
D
M

ZKLWHýSDSHU
6
+3ý1HW6HUYHUý$FFHVVRULHV
Backup
System backup is an essential component of any mass storage strategy. Backup strategies
provide for the reconstruction and recovery of lost operating environments and application
data if the system should fail. They complement other availability strategies and are
designed to improve the overall system uptime.
Backup is a combination of hardware, software, and required operator actions. Automated
backup is an option and is recommended for most installations. A wide variety of tape
backup choices, including QIC/Travan, DAT, DLT, autoloaders, and tape libraries, are
available, and consumers make their selections based on the reliability, capacity, transfer
rate and, above all, value for the money of these choices. Certain RAID structures,
including duplexing and mirroring, can supplement backup, since copies of the data are
available to ensure reproducibility, but RAID is not designed to be a substitute for backup
routines.
Each server environment should have a well-documented plan for backing up important
data. It is important to apply a “system-level” backup strategy, not just a file-level or
application-level backup strategy. Indeed, a Disaster Recovery (DR) approach to backup is
increasing in popularity. This approach ensures that in the event of a total system failure,
when not even the operating system can recover, a backup tape can completely restore the
server with its operating system, configuration files, applications and user data.
Backup management is an important consideration as well. Management tools are
available that monitor the backup process so that the system administrator becomes aware
of any error conditions that might arise during the backup operation. The more
sophisticated tools also advise a course of action that corrects the error. These tools help
to ensure that backups are successfully completed.
Users continue to demand higher levels of availability; minimizing downtime is essential.
Backup software solutions are now available which permit backups to occur while the
server is available to users. Some hardware configurations allow tape drives to be replaced
“on-line” without powering down the server.
Finally as space becomes a premium in many IT environments, backup products are now
designed to be rack-mountable as well as stand-alone.
Issues to Consider
This discussion should lead the reader to a set of fundamental questions about server data
protection and your server mass storage solutions.
Does the server vendor really understand mass storage?
Does the company have a track record supplying systems into demanding enterprise
environments, or is it a PC company trying to address server needs with mass storage
designed for desktop use?

ZKLWHýSDSHU
7
+3ý1HW6HUYHUý$FFHVVRULHV
Is the server’s mass storage protection integrated with its management software?
Since electromechanical devices like drives are more likely to fail than other parts of my
server, can the server management system—locally or remotely—completely view and
control the disk subsystem?
Does the vendor’s storage performance meet expectations?
Do the disks provide sufficient capacity? Are the data transfer rates fast enough?
How should data and applications be organized on the server?
Are server usage patterns understood?
Are adequate backup procedures in place for important server data?
The data protection techniques described in this paper can be used to make server disk
subsystems extremely reliable, but backup is still absolutely necessary. These techniques
do not protect from accidental or malicious destruction, theft, or site-wide problems like
flood or fire.
What kind of data protection is appropriate?
Is it acceptable to rely on disk backup procedures alone, replacing drives when they fail,
temporarily interrupting users and possibly losing new data not yet backed up? Is it
acceptable to continue operating, but with degraded performance, after a failure? Or do we
require a solution that ensures uninterrupted, full-speed performance despite failures?
What is the cost of downtime?
Can the cost to the organization be expressed in financial terms like lost revenue or profit
per hour or day, allowing an intelligent decision about data protection? How closely related
is this server to revenue or profitability, or to the productivity of the employees?
HP NetServer Solutions
Hewlett-Packard has been supplying system solutions for demanding enterprise
environments for over 20 years. This experience gives HP unmatched expertise in
understanding and solving the needs you have for mass storage.
As a technology leader and as a developer and supplier of disk drives, tape drives, and
mass storage subsystems, HP has a wealth of knowledge about mass storage. As one of
the top suppliers worldwide of server systems, HP has also developed close relationships
with the leading drive suppliers, making sure they understand HP’s requirements and
helping them make their products available to you on a timely basis.
The following sections show some of the important ideas behind HP’s NetServer mass
storage product offering.

ZKLWHýSDSHU
8
+3ý1HW6HUYHUý$FFHVVRULHV
Hard Disk Drives
SCSI hard disk drives are available in a variety of capacities from 4 to 18 GB, with higher
capacity drives (36 GB) planned for mid-1999. HP offers drives at two rotational speeds:
7200 and 10,000 rpm. The higher speed drive offers shorter seek times and improved
system performance.
Drives are available for either internal or external storage systems. HP offers two versions
of external storage: rack-optimized or pedestal/standalone configurations.
HP’s position as a leading server vendor gives the company considerable leverage with
leading drive vendors. HP has close relationships with these drive vendors, allowing HP to
work with them on new drive technologies and standards, to thoroughly test these
technologies with HP NetServer hardware and software, and to bring them to market in
cost-effective and timely ways. As an example, HP offers proactive warranty replacement
on drives, meaning that the company will replace them for you if HP server management
software indicates a problem—even if a failure has not yet occurred. Current areas of
cooperation with drive vendors include both higher capacity and faster drives, enhanced
drive connection mechanisms, and better availability and management features.
Performance improvements include 10K RPM drives and low voltage differential (Ultra2)
SCSI technology.
HP NetServers offers standard SCSI controllers in all server systems. Additional SCSI,
Disk Array, or Fibre Channel controllers may be added. The HP NetServer website
provides lists of tested controller cards, organized by platform, as well as more information
about each of the controller products.
HP NetServer RAID Configurations
RAID can be implemented in software or hardware. Software-based RAID uses host CPU
cycles and system memory adding processing overhead that may impact system
performance. Disk array controllers (DAC), the hardware solution, transfer the overhead
from the host CPU to specialized hardware that typically performs better than the software
RAID.
The DAC can be programmed to support a
hot spare drive configuration. With this
arrangement, if a drive should fail, the
system can automatically begin rebuilding
to the hot spare drive. See Figure 4 for an
explanation of the differences between a
hot swap and a hot spare configuration.
There are several RAID levels to choose
from whether the implementation be
with software or hardware. Five levels
(1 through 5) were defined when the RAID
concept was first proposed in 1988 by
three academics from the University of California at Berkeley as a way to organize and
protect stored data. Additional levels have since been developed to enhance performance
and data redundancy. Hewlett-Packard NetRAID family controllers offer a choice of seven
XHot Spare
Hot Swap array controller
rebuilds data
bad drive is
replaced
Figure 4. Hot Swap versus Hot Spare

ZKLWHýSDSHU
9
+3ý1HW6HUYHUý$FFHVVRULHV
different levels (0, 1, 3, 5, 10, 30 and 50) which allow users to vary the degrees of
performance, availability and flexibility.
I2O
The I2O (Intelligent Input/Output) specification defines an architecture for intelligent I/O that
is independent of both the specified device being controlled and the host operating system
(OS). The I2O specification has been developed to address two key problem areas in
I/O processing:
•System performance degradation caused by I/O interrupts to the CPU (the CPU is not
doing data processing when it is processing I/O requests)
•The necessity to create, test, and support unique drivers for every combination of
I/O device and OS on the market
HP was the founding company of the Special Interest Group (I2O SIG) and has had a
significant input into the direction of the development effort of the specifications, including
the initial invention, and has been assigned the patent.
The new architecture of I2O utilizes an approach whereby I/O processors are used to
handle I/O rather than the CPU which offloads low-level interrupts to these processors. The
I2O specification also creates a “split driver” model so that the software drivers are portable
across multiple OSs and host platforms. One portion of the driver runs in the operating
system, the other portion runs in a separate I/O system. The split driver concept will reduce
the number of drivers that must be written. OS vendors will write a single I2O-ready driver
for each class of device—such as a disk drive—and the device manufacturers write a
single I2O-ready driver for each device, which will work for any OS that supports I2O. The
split-driver approach will drive down the cost of software development and support for
I/O devices while increasing their availability across the range of operating systems.
Other benefits that will come as I2O is fully implemented are improved server efficiency,
reduced CPU utilization, “plug-and-play” flexibility, and a more stable I/O driver platform.
I2O is being implemented step-by-step like many technology standards that came before it.
Operating system vendors began releasing the systems that provide support for I2O in early
1998. Hewlett-Packard has launched the HP NetRAID-3Si disk array controller, the first in a
line of HP NetServer products that will support the I2O standard.
Fibre Channel
Just as I2O capability improves the input/output transfers and the performance of the
entire server system, the addition of Fibre Channel connectivity enhances system
performance and availability. Fibre Channel was first offered with HP-UX server storage
and has become widely available with NT-based servers over the last year. There are
currently 3 external disk storage products that can be connected to HP NetServers via
Fibre Channel—the HP Rack Storage/12FC, the HP Model 30/FC and the EMC Symmetrix.
HP also offers connectivity of DLT Backup Libraries to FC via the HP SureStore Fibre
Channel SCSI Bridge.
Fibre Channel connections between server and storage are made using a set of
components similar to those used for IP networking. A Fibre Channel-to-PCI host bus

ZKLWHýSDSHU
10
+3ý1HW6HUYHUý$FFHVVRULHV
adapter (HBA) facilitates point-to-point connections. The addition of a hub changes the
Fibre Channel topology from point-to-point to arbitrated loop (FC-AL) and increases
configuration flexibility. Multiple servers and storage units can be interconnected using
HBAs and one or more Fibre Channel hubs. The availability of individual and clustered
servers can also be enhanced when hubs are used to build redundancy into configurations.
Fibre Channel switches support a third topology, fabric, and are becoming widely available
now. The word fabric is meant to convey the image that multiple connections can be
interwoven through the switch, in contrast to the loop orientation of a hub. Unlike hubs,
switches also scale bandwidth upward with each connection made and have the
intelligence to handle certain processes and organize ports into exclusive groups called
zones. Fibre Channel HBAs, hubs, and switches make it possible to build storage networks
that can be managed much like IP networks. Over the next few years, new software will
make it possible to treat storage as a utility that is shared between servers.
Backup Solutions
HP offers a full range of hardware and software products used to back up data on the
server system. For the entry-level servers, QIC/Travan offers cost-effective data protection.
Digital Audio Tape (DAT) provides high performance, high capacity, low cost of ownership,
industry-standard compatibility, and a well-established upgrade path. Digital Linear Tape
(DLT) offers high-end performance for business-critical applications, especially for those
systems where the total disk capacity is greater than
40 GB. High capacity disks require rapid backup solutions. HP offers a rack-mountable,
high availability, tape backup product and TapeRAID backup application software that
stripes backup data across several tape drives. With this configuration, it is possible to
back up almost 1 GB of data per minute.
In the Field of Disaster Recovery solutions, HP offers the world’s first hardware based One
Button Disaster Recovery process (OBDR) for Intel Servers, with its HP SureStore DAT
drives. When used with OBDR compatible software this feature saves the DR “boot image”
onto tape as part of your daily backup. Thus the configuration data pertaining to your
system is always kept current and stored with the backup on tape. Even if a total system
failure should occur, the system can still be easily recovered by pressing just one button on
the DAT drive, whereupon the DAT drive emulates a CDROM to recover your total system
from the latest backup tape(s).
Automated backup solutions are strongly recommended. HP SureStore Autoloaders and
DLT Libraries add reliability to the backup process, since the backup is always performed
as scheduled onto the correct cartridge. Automation also frees the IT staff for more
demanding duties and, in remote locations where IT support may not be available, ensures
that the backup is performed.
Clustered server systems increase backup complexity, since the data is actually shared by
2 or more “nodes” in the cluster. The ideal solution to back up Windows NT clusters is to
have the backup devices shared by the nodes of the cluster. The best medium for sharing
devices is Fibre Channel. In this domain, tape libraries become the dominant solution,
offering high capacity, automation and flexibility in backup scheduling. HP currently offers a
HP SureStore Fibre Channel SCSI Bridge that allows multiple servers to be backed up to a
single device via FC. The bridge also enables the connection from SCSI DLT Libraries to
FC storage area networks.

ZKLWHýSDSHU
11
+3ý1HW6HUYHUý$FFHVVRULHV
Integrated Management
HP NetServer disk subsystems are completely integrated with Hewlett-Packard server
management software, allowing you to monitor and manage disks on local and remote
servers. For example, if you have installed HP TopTools for Servers, then software agents
automatically monitor the server system and provide the following server mass storage
alerts to the system operator:
•Disk capacity threshold alert and usage tracking
•SCSI controller and device failure notification (co-developed with Adaptec®)
•Predictive failure analysis that warns of probable disk problem(s)
HP SureStore tape backup devices support the “TapeAlert” feature that makes backups
more manageable. TapeAlert functions as a continuous monitor that informs backup
applications of possible erroneous conditions that could be developing in any backup
device. Further, it specifies the appropriate course of action to the backup administrator.
Thereby, TapeAlert helps to assure that all backups are completed successfully.
Summary
As we have demonstrated throughout this paper, HP has continually provided leadership in
the areas of storage and controller development and product implementation. Most
recently, HP has been very active with the I2O and Fibre Channel industry groups and is
now offering leading-edge products with these capabilities.
As you, the reader, carefully consider your storage needs from the viewpoints of the issues
and solutions presented here, it is clear that HP has the experience to understand your
mass storage needs. HP will continue to offer products, based upon the best choices in
technology, that will meet your performance, availability, and manageability needs both
today and tomorrow.

ZKLWHýSDSHU
12
+3ý1HW6HUYHUý$FFHVVRULHV
For More Information
For additional information about single server and multi-server high availability see several
papers at:
http://www.hp.com/netserver/products/cluster/more.htm
For additional information about data protection schemes, see the RAID Levels Technology
Brief at: http://www.hp.com/go/netserver/pdf/raid.pdf
For additional information about tape devices, Disaster Recovery, TapeAlert and high
availability backup strategies, please consult the relevant papers located at
http://www.hp.com/go/tape.
For additional information about HP Fibre Channel disk products, see:
http://www.hp.com/netserver/products/rs12
For performance information of tape devices, see:
http://www.hp.com/netserver/pdf/sql_dbbackup.pdf
For additional information regarding the performance of 10K disk drives, see:
http://www.hp.com/netserver/pdf/diskcmp.pdf
For additional information about common tray disk drives, see:
http://www.hp.com/netserver/products/accessories/hdd_curr.htm
For additional information about hot swap disk drives, see:
http://www.hp.com/netserver/products/accessories/hdd_curr2.htm
Microsoft Cluster Server is a registered trademark of Microsoft Corporation. Microsoft Windows NT is a registered trademark of Microsoft Corporation.
All other brand names are trademarks of their respective owners. Information in this document is subject to change without notice.
©Copyright Hewlett-Packard Company 1999. All Rights Reserved. Technical information in this document is subject to change without notice.