product

HP
Data Center
HyperPlex

HP 9000

Data Center Manageability, Extreme Capacity

The growth of distributed computing technology during the 1990s
has enabled the information technology (IT) industry to fill its data
centers with powerful server nodes, capable of processing and
transferring data at unprecedented speeds. Such growth has its
benefits; however, the downside is that these extensive resources
can quickly become unwieldy, requiring ever-larger
staffs to manage and keep the servers up and running.
The answer? The mainframe-class manageability and power of
Hewlett-Packards HyperPlex Data Center Solutions.
HPs HyperPlex platform is an integrated solution combining HPs
industry-leading HP 9000 Enterprise Server technology, HPs new
ServiceControl central management suite, and HP high-speed
server interconnects. HyperPlex assures peak UNIX application
performance and mission-critical high availability, necessities in
todays IT environments. The extreme compute power and capacity
of HyperPlex is required to meet the rigorous demands of data
center systems consolidation, enterprise resource planning (ERP),
data warehousing, and online transaction processing (OLTP), as
well as large-scale, highly complex technical modeling and
simulations.
The result is what every IT manager wants out of a data center
platform
the power to reduce complexity and increase productivity while
meeting the growing demands of business objectives within their
enterprise data centers.

HP HyperPlex and HP ServiceControl:

Taking Back Control of the Data Center
Complex multi-server data centers can quickly become a burden
when excessive time and resources are required to manage them
or worse, when they are not being used to their full advantage. In
response, HPs HyperPlex platform incorporates HP ServiceControl,
a systems management suite that enables IT managers to run their
high-performance UNIX-based environment with the discipline of a
mainframe solution.

HP Central Web Console

At the heart of HP ServiceControl is the Central Web Console,
designed to ease the management of large data center clusters.
Previously, administration of systems has been dictated by
geography, requiring administrators to be present at each location.
Now, using Web-browser access through the Central Web Console,
system administrators can manage large system clusters
from anywhere in the company or in the worldsaving time and
expenses. Connections between the Web browser and HyperPlex

The Benefits of HPs HyperPlex

Solution
Central control of data center
compute resources, reduced
overhead expenses, and high
availability assurance with
HPs powerful ServiceControl
systems management suite
State-of-the-art capacity of
up to
64 highly scalable individual
nodes (thousands of CPUs)
interconnected
with ultra-high bandwidth,
low-latency HP HyperFabric
networking interconnect
Flexibility to increase
HyperPlex compute capacity
through the
dynamic addition of server
nodes
and system resources
Rolling version upgrades of
existing nodes, CPUs, and
operating system instances
all without disruption to
the production environment
Investment protection and
extended platform life with

servers are completely secure, password-protected, and functional

within firewall environments.

Data Center Manageability, Extreme Capacity

Robust Systems Management

Software
Using the robust systems management software
of
HP ServiceControl, data center administrators are
able
to view the status of the HP HyperPlex
environment
as well as see out across the integrated,
networked
IT infrastructure. Using drill-down capabilities,
the administrator can view the status of each
HyperPlex node, along with the application
package installed on
each one. With a centralized platform, IT
managers can allocate applications and
associated resources across
the HP HyperPlex environment to ensure specific
service level objectives (SLOs) are met.
For example, as a function of the central
management
of HP HyperPlex, IT managers can dynamically
change system CPU, memory, and I/O resources
to ensure that the most critical applications get
the resources they need. Servers, storage
devices, and applications within the system can
be added, removed, and modified without
impacting overall system productivity. With HPs
MC/ServiceGuard, an integral part of HP
ServiceControl, system administrators can also
ensure high availability and system uptime for
every mission-critical application. The result: the
HP HyperPlex platform transforms the data center
from an untamed powerhouse to a finely tuned,
easily managed system, programmed to run
more cost-effectively and productively around the
clock.

Performance Capacity Thats Real

HP HyperPlex solutions provide IT managers with
the ability to expand their computing capabilities
beyond those available in a single symmetric
multiprocessing (SMP) system. Using HPs ultrahigh-speed, low-latency HyperFabric server node
interconnect, up to 64 high-performance HP 9000
Enterprise Server nodes can be linked. Each of
these nodes can have as many as 128 processors
in the case of the V2500 Scalable Computing
Architecture (SCA) configuration. As a result,
HyperPlex can theoretically deliver an
unprecedented 8,192 powerful PA-RISC CPUs to
power the most extreme computing task. Future
upgrades in server nodes and in
the HP HyperFabric interconnect will support still
more processorsdemonstrating the unlimited
growth potential of HPs HyperPlex environment
and its ability
to size up and stay ahead of the most demanding
IT requirements. In addition, a series of future
performance enhancements for the HP 9000
HyperPlex platformincluding future versions of
the 64-bit PA-8xxx family
and the IA-64 processor within server nodeswill
offer
still higher performance and increased system
consolidation options.

Data Center Manageability, Extreme Capacity

Features of HP HyperPlex
Central management and availability
HP ServiceControl provides Web-based
administration, application control, resource
management, and high availability software to
control the data center.
Extreme capacitycapability and capacity
far exceeding that of a single SMP system.
High-speed connectivityimplements fast,
flexible connections using HPs HyperFabric
high-bandwidth, low-latency node
interconnect, as well as other traditional
networking technologies such as FDDI,
Fast Ethernet, Gigabit Ethernet, and
asynchronous transfer mode (ATM).
Mainframe UNIX performanceachieves
the
highest possible application performance, in
excess
of mainframe levels, to meet the needs of the
most demanding UNIX applications. Workload
is
balanced across powerful computing nodes
within
the HP HyperPlex environment, maximizing
throughput efficiency of overall system
resources.
Rolling upgrade flexibilitypermits
upgrades through multiple generations of
systems, processor technology, and operating
system versions, ensuring investment
protection.
Optimized use of floor spacecontributes
greatly
to reduced cost of ownership. HP HyperPlex
solutions allow centralization of maintenance
terminals and dense data center racking.
Systems consolidationreduces the number
of
servers in a burgeoning IT environment while
improving service levels and lowering costs.
Systems consolidation greatly simplifies
administration functions, increases security,
and lowers the
overhead costs of an IT environment.

Mass storage consolidationincreases

performance and availability using industryleading Fibre Channelbased mass storage
devices.
Backup consolidationprovides backup
capacity in excess of 1 TB/hour using HP
HyperFabric.
Factory integrationdelivers HPs
ServiceControl management suite and all
system hardware resources fully pre-integrated
for faster, smoother deployments.

Multiple Avenues for Systems

Consolidation
As an IT organization brings more and more
applications online, the tendency is to acquire
more and more servers to meet the additional
demand of the new applications. Yet it is vital to
keep the time and effort required to manage
these increasingly capable but complex systems
to an absolute minimumespecially given the
budget constraints IT professionals face in
todays competitive business environment.
HP HyperPlex simplifies an IT environment in
multiple ways. For example, administrators can
adjust and allocate capacity (CPU, memory, mass
storage, and backup) by adding, reducing, or
redeploying resources as demanded by the
application environment. This allows optimized
use of resources, applied exactly to the area
where
they are most needed and best utilized. In
addition,
the streamlined nature of HP HyperPlex can
potentially reduce the total number of servers
and mass storage peripherals required in a
client/server environment. The remaining servers
and peripheral subsystems can then be centrally

managed more efficiently, resulting in greatly

reduced cost of ownership along with increased
productivity of individual users and workgroups.

Data Center Manageability, Extreme Capacity

HP-UXthe Enterprise Operating

Environment
of Choice
The foundation for HP HyperPlex solutions is the
industry-proven HP-UX operating environment.
The Hewlett-Packard UNIX (HP-UX) operating
system provides the standards-based, robust,
time-tested 64-bit operating environment
required for high-performance, mission-critical
applications. Running the HP-UX enterprise-class
operating system, the HP 9000
HyperPlex solutions are compatible with the HP
9000 Enterprise Server product family. This
seamless product compatibility supports more
than 14,000 off-the-shelf commercial and
technical HP-UX applications. Using
HP-UX 11.0, the D-, R-, K-, T-, and V-Class server
node solutions offer an unmatched number of
supported UNIX application choices.

Industry-Leading High Availability

with
HP MC/ServiceGuard
HP MC/ServiceGuard, an element of the
HP ServiceControl systems management suite, is
the industry-leading high availability (HA) solution
for the open systems environment designed to
protect mission-critical applications from
hardware and software failures.
With HP MC/ServiceGuard, multiple server nodes
in
an HP HyperPlex platform are organized into an
HA environment capable of delivering highly
available application performance to LANattached clients. In conjunction with Event
Monitoring Service (EMS),
HPs MC/ServiceGuard monitors the health of
various components within each server node and
quickly responds to failures in a way that
eliminates or minimizes application downtime.
MC/ServiceGuard is able to detect and respond to
failures in:
CPUs
System memory
System and application processes
LAN media and adapters
High-speed node interconnect
Mass storage interconnect and subsystems
Applications

For ease of management and outstanding

flexibility,
HP MC/ServiceGuard allows system
administrators to organize all the resources
needed by an applicationsuch as disks, network
resources, and application or system processes
to be organized into application packages.
These application packages are managed and
moved within the HP HyperPlex environment. HP
HyperPlex solutions permit flexibility in designing
HA topologies, from mirrored (totally redundant)
components to active/active or active/standby
configurations. Additionally, HP delivers a pretested, pre-configured package of HP hardware,
software, and support services. This high
availability solution, called Mission Critical Server
Suites (MCSS), delivers a 99.95% uptime
commitment.
When HPs High Availability Foundation
configuration management is combined with a
Critical Systems Support (CSS) contract, HP will
guarantee customers better than 99.95%
application availability.

Multiple High-Speed Interconnect

Options
High capacity provides the growth path, and highperformance connectivity solutionsfor both
system-to-system and system-to-mass-storage
networkingallow the HP HyperPlex platform to
perform like one large mainframe-class system.
Keeping pace with IT managers ability to scale
their systems to record-high levels, HP offers
ever-increasing flexibility of choice. HewlettPackard connectivity products include
HP HyperFabric, Gigabit Ethernet, FDDI (Fibre
Distributed Data Interface), Ethernet, and
asynchronous transfer mode (ATM) for server
node interconnects,
as well as Fibre Channel for shared mass-storage
connectivity, all of which can be fully integrated
at
the HP factory.

HP HyperFabric: Ultra-HighPerformance, High Availability

Interconnect
Offering both high performance and excellent
flexibility, HP HyperFabric is a high-speed, low-

Data Center Manageability, Extreme Capacity

latency virtual backplane interconnect that runs

standard TCP/IP protocols and HPs low-latency
protocols for commercial and technical
application scaling. The HyperFabric switch
consists of 16 communication ports, each of
which can independently communicate at a
speed of 2.5 Gb/s. Meshing of switches is
supported up to configurations
of 64 server nodes.
And for applications that demand still faster
communication, simply adding another HP
HyperFabric interface per node will automatically
double performancewith no required application
changes
or complex network routing issues. For instance,
up to
8 HyperFabric interface cards can be installed in
each
V-Class system. Furthermore, HyperFabric does
not require an idle redundant interface for local
network failover.
HP HyperFabric interconnect, comprising
input/output (I/O) interfaces, switches, and
cabling, can be used in either point-to-point or
switched configurations. High availability is
provided through the use of active/active
connections in the point-to-point mode and
through redundant switches in the switched
mode. The number
of I/O slots needed for HyperFabric, compared to
FDDI,
is greatly reduced.
The benefits of using HP HyperFabric interconnect
in
the data center extend across the most
demanding applications. Using HyperFabric in the
HP HyperPlex environment yields an increased
node-to-node throughput, providing enhanced
performance levels to meet the large-scale
requirements of implementations such as data
warehousing and ERP. Data backups for these and
other large applications can be performed at
speeds of greater than 1 Tbyte/hourbeating
that of average networking technologies by a
wide margin.

HP HyperPlex Protects Your IT

Investment
Traditional mainframes or other proprietary
solutions
can lock customers into years of hardware and
software investment. HP HyperPlex platform
solutions offer future reinvestment options, such
as the flexibility to introduce new platform
technology or redeploy powerful server nodes as
file, application, or database servers.

The benefit of both upward and downward

adjustments
to HP HyperPlex, plus the ability to consolidate
distributed computing resources, is a level of
system investment protection unsurpassed by
any other type
of enterprise computing resource. This is
especially important today, given the frequency
of new platform
and processor technology introductions. For
example,
as HP introduces Intels IA-64 processor
technology,
these new systems can be smoothly integrated
into the HP HyperPlex solution environment.
Further reductions in IT and research and
development budgets can be realized by rolling
upgrades of application and system software
into the HP HyperPlex solution. This capability
allows IT managers to make a smooth transition
to new software technologywith minimal impact
on the production environment.

Flexible HP HyperPlex Architecture Is

Ready
for Growth
The HP 9000 HyperPlex solution integrates a
combination of powerful HP 9000 symmetric
multiprocessing (SMP) server nodes, high-speed
interconnects, and clustered system
management tools. SMP functionality can be
augmented by adding V-, K-, D-, R-, or T-Class
systems, and combinations of these server nodes
can be interconnected using HPs HyperFabric or
other traditional networking technology. Finally,
mass storage can be connected using industryleading, multi-system Fibre Channel technologies.
Building up the number
of SMP nodes in this way enables a linear
increase in performance capacity, I/O bandwidth,
and other system resources such as memory,
mass storage, or compute capacity.

Data Center Manageability, Extreme Capacity

HP 9000 HyperPlex Data Center

Solution
With its unique architecture, HPs HyperPlex is a
flexible and scalable platform environment that
can easily grow
to accommodate large-scale data center or
enterprise consolidation, parallel computing
performance requirements, and mission-critical
high availability. Emerging technologies such as
parallel database management solutions, highspeed switching, and future servers based on the
highly anticipated Intel IA-64 processor
technology can be added as needed. In short,
HPs HyperPlex platform brings together the best
enterprise computing products HP has to offer,
for a solution set that can be tailored to meet a
customers largest computing requirements.

Technical Application Development

Tools
To increase application performance on a single
job beyond the level achievable in a single
processor or SMP node, HP offers the MessagePassing Interface (MPI).
MPI is the leading standard library for
implementing applications using the messagepassing model. HP has implemented MPI to allow
application developers and users to take full
advantage of the HP HyperPlex
platform capabilities.

Parallel Development Tools

For parallel technical applications, HP provides
automatic parallelizing compilers, a parallel
performance analyzer, and an optimized
mathematical library. These tools together create
a powerful, straightforward programming
environment that helps the user create some of
the most efficient parallel code possible. The
environment supports shared-memory, message
passing, and hybrid approaches to parallel
programming. This powerful set of tools provides
a high-productivity environment that eases the
development and porting of highly parallel
applications.

HP Delivers a Complete Solution

Many organizations today look for a turnkey
solution
to data center computing problems. Building such
an environment that adapts to real-world
computing situations consumes precious
company time and resources. HP solves this
problem in the factory by predefining a large
number of system templates that include
management software, HP 9000 server nodes,
interconnects, mass storage, and cabinets. This
level
of integration ensures that each system is
packaged as
a complete solution, ready to install in a data
center environment, and backed by HPs worldclass support organization.
Should one of the pre-configured templates not
suit a customers needs, HP technicians can
utilize
a sophisticated configuration and analysis tool to
customize an HP HyperPlex configuration. Along
with this customization capability, HP offers an
array of enterprise-class, mission-critical
technology delivery programs to ensure that
every HP 9000 HyperPlex solution is properly
sized and integrated into the customer
environment.
Leading the industry, HP is committed to
providing not just some, but all elements of a
complete data center solution, including best-inclass technology, services, consulting, finance,
support, and licensing solutions.

Data Center Manageability, Extreme Capacity

HP HyperPlex Single-SMP Server Node

Performance Specifications

D-Class

R-Class

K-Class

CPUs per engine (max.)

2-way SMP

2-way SMP

6-way SMP

Standard memory
Maximum memory

64 MB
3 GB

128 MB
3 GB

128 MB
8 GB

Maximum disk capacity per engine

Processor type

9.5 TB
64-bit PA8200
2 MB/2 MB
240 MHz

9.5 TB
64-bit PA8200
2 MB/2 MB
240 MHz

35.5 TB
64-bit PA8200
2 MB/2 MB
240 MHz

Instruction/data cache
Clock speed

T-Class
(T-600)
12-way
SMP
256 MB
3.75 or 32
GB1
>40 TB
64-bit PA8200
1 MB/1 MB
180 MHz

V-Class
128-way
SMP
4 GB
128 GB
>160 TB
64-bit PA8200
MB/1 MB
440 MHz

HP HyperPlex Multi-SMP Server Node Specifications (with HyperFabric

Interconnect)

Maximum total engines (engines of any model)

Interconnect Specifications Ethernet
(10/100T) FDDI Fibre Channel/266
HyperFabric
Peak bandwidth (Mbits/s)
Latency (microseconds)
1
2
3
4

HP-UX 11.0 64-bit

Using TCP-Lite protocol
TCP/IP
Utilizing future software release

D-Class/
R-Class
64

K-Class

T-Class

V-Class

64

64

64

10/100
250 s

100
250 s

266
125 s2

2,560
<100 s3/
<30 s4

For the location of the nearest sales office call:

United States of America: +1 800 637 7740
Canada:
Hewlett-Packard Ltd.
5150 Spectrum Way
Mississauga, Ontario L4W 5G1
+1 905 206 4725
Japan:
Hewlett-Packard Japan, Ltd.
Japan Country H.Q.
3-29-21, Takaido-Higashi, Suginami-ku,
Tokyo, 160-8585 Japan
+81 3 3331 6111
Latin America:
Hewlett-Packard
Latin American Region Headquarters
Waterford Building, 9th Floor
5200 Blue Lagoon Drive
Miami, Florida 33126 USA
+1 305 267 4220
Refer to country phone numbers
Australia/New Zealand:
Hewlett-Packard Australia Ltd.
31-41 Joseph Street
Blackburn, Victoria 3130
Australia (A.C.N. 004 394 763)
+61 3 9272 2895
Asia Pacific:
Hewlett-Packard Asia Pacific Ltd.
17-21/F, Shell Tower
Times Square
1 Matheson Street
Causeway Bay
Hong Kong
+8522 599 7777
Europe/Africa/Middle East:
Hewlett-Packard S.A.
150, Route du Nant-dAvril
CH-1217 Meyrin 2
Geneva, Switzerland
+41 22 780 81 11
European Multicountry: +41 22 780 81 11
Middle East and Africa: +41 22 780 71 11
European Headquarters: +41 22 780 81 81
Refer to country phone numbers

For direct country contact call:

Argentina: +541 787 7145
Austria: +43 1 25 000 0
Belgium and Luxembourg: +32 2 778 31 11
Brazil: +5511 7296 8000
Chile: +562 203 3233
Colombia: +571 629 5030
Denmark: +45 45 99 10 00
East Central Europe, CIS, and Yugoslavia:
+43 1 25 000 0
Finland: +358 9 887 21
France: +33 1 69 82 60 60
Germany: +49 7031 140
Greece: +30 1 689 644
Hungary: +36 1 252 7300
Iceland: High Performance Systems hf. +354 1 67
10 00
Ireland: +353 12 88 33 99
Israel: Computation and Measurement Systems
(CMS) Ltd. +972 3 5380 333
Italy: +39 2 92122770
Mexico: +525 326 4600
Netherlands: +31 20 547 6911
Norway: +47 22 7356 00
Poland: +48 22 608 77 00
Portugal: +351 1301 7343
Russia and the CIS, excl. Ukraine: +7 095 923
5001
Slovenia: +38 61 55 84 72
Spain: +34 1 631 1600
Sweden: +46 8 444 2000
Switzerland: +411 735 7111
South Africa: Hewlett-Packard
South Africa (Pty) Ltd. +27 11 806 1000
Turkey: +90 212 224 5925
United Kingdom: +44 1344 369231
Venezuela: +582 239 4133

Or check out our Web sites at

www.hp.com/go/datacenter or
www.hp.com/go/hp9000
Intel is a U.S. registered trademark of Intel Corporation.
UNIX is a registered trademark of The Open Group.
Technical information contained in this document is subject
to change without notice.
Copyright Hewlett-Packard Company 1999
All Rights Reserved. Reproduction, adaptation, or translation
without prior written permission is prohibited except as
allowed under the copyright laws.

Printed in USA 03/99

5968-4856EUC