• Application node and control node options expanded according to latest HP ProLiant server offering.
• Includes new HP ProLiant DL165 G5 server.
• Support for HP ProLiant BLxc220c, BL260c G5,BL465c G5, BL685c G5, DL385 G5 and DL585 G5 servers
  

• One cabinet: up to 32 compute nodes with Myrinet, Gigabit Ethernet, InfiniBand and Quadrics Interconnects
• Two cabinets: up to 64 compute nodes
• Three cabinets: up to 96 compute nodes
• Four cabinets: up to 128 compute nodes
  
  Maximums based on selecting 1U compute nodes and no utility nodes. Selection of larger nodes (e.g., HP ProLiant DL585 or HP ProLiant DL380) or MSA Storage option will reduce node count per cabinet.

Each dense cluster includes one Utility Building Block (UBB) cabinet, which houses the control node, RKM, and the optional high-performance interconnect switch, and utility nodes. Other cabinets are used to house the compute nodes. Compute nodes may be integrated into the UBB, as space permits, in dense configurations.

HP Cluster Platform modular solutions are organized in building blocks that can be easily replicated to build large-scale clusters.

• The compute building block (CBB) is a cabinet that contains, among other things, up to 38 2-socket 1U compute nodes, up to nine 4U compute nodes or up to four c-Class enclosures. One enclosure can house up to sixteen half-height blade nodes or up to eight full-height blade nodes.
• The utility building block (UBB) is the core cabinet of the cluster that contains the control node and its keyboard/monitor, additional utility nodes if required, cluster management root switches, one optional high-speed interconnect switch for solutions up to 128 nodes and optional MSA1500 or MSA60 raid arrays.
• The utility eXpansion rack (UXR) is an optional cabinet that can contain up to ten utility nodes when no more room is available in the UBB.
• The interconnect building block (IBB) is a cabinet that contains multiple interconnect switches for solutions that are larger than 128 nodes.
• The visualization building block (VBB) is a cabinet that contains up to 8 5U HP personal workstation visualization nodes.
• Smallest rackmount server configuration: one UBB cabinet and one CBB cabinet-up to 38 compute nodes
  
• Smallest blade server configuration: one UBB cabinet with up to 3 c-Class enclosures
  
• Maximum number of CBB: 35 for rackmount servers, 22 for blade servers
  
• Maximum number of UBB: 1
• Maximum number of UXR: 1
• Maximum number of IBB: 3 with Myrinet, 1 with Gigabit Ethernet, 4 with InfiniBand, 6 with Quadrics
• Maximum number of VBB: 12

HP ProLiant DL380 G5 Server (recommended for PCI slot availability)
HP ProLiant DL160 G5 Server
NOTE: Complete information on the HP ProLiant Servers can be found in the following QuickSpecs:
http://h18000.www1.hp.com/products/quickspecs/Division/10480.html

All currently shipping processors for each control node are available. Select one or two processors

HP ProLiant DL160 G5 Server
Configure servers as base for one or two compute pools. Total nodes cannot exceed 512 nodes.
NOTE: Complete information on the servers can be found in the following QuickSpecs:
http://h18000.www1.hp.com/products/quickspecs/12509_div/
12509_div.html

All currently shipping processors for each compute node are available. Select one or two processors.

HP ProLiant DL380 G5 Server (recommended for PCI slot availability)
HP ProLiant DL160c G5 Server
HP ProLiant BL260c G5
HP ProLiant BL460c
HP ProLiant BL480c
NOTE: Complete information on the HP ProLiant Servers can be found in the following QuickSpecs:
http://h18000.www1.hp.com/products/quickspecs/division/10455.html

All currently shipping processors for each control node are available. Select one or two processors

HP ProLiant BL260c G5
HP ProLiant BL460c
HP ProLiant BL480c
Configure servers as base for one or two compute pools. Total nodes cannot exceed 512 nodes.
NOTE: Complete information on the servers can be found in the following QuickSpecs:
http://h18000.www1.hp.com/products/quickspecs/division/10455.html

All currently shipping processors for each utility node are available.

NOTE: Complete information on the servers can be found in the following QuickSpecs:
http://h18000.www1.hp.com/products/quickspecs/division/10455.html

HP ProLiant DL385 G5 Server (recommended for PCI slot availability)
HP ProLiant DL165 G5 Server, HP ProLiant DL585 G5 Server
NOTE: Complete information on the HP ProLiant Servers can be found in the following QuickSpecs:
http://h18000.www1.hp.com/products/quickspecs/division/10480.html

All currently shipping processors for each control node are available. Select one or two processors (two or four with HP ProLiant DL585)

HP ProLiant DL585 G5 Server
HP ProLiant DL165 G5 Server
Configure servers as base for one or two compute pools. Total nodes cannot exceed 512 nodes.
NOTE: Complete information on the servers can be found in the following QuickSpecs:
http://h18000.www1.hp.com/products/quickspecs/Division/10480.html

All currently shipping processors for each compute node are available. Select one or two processors (two or four with HP ProLiant DL585)

HP ProLiant DL385 G5 Server (recommended for PCI slot availability)
HP ProLiant DL165 G5 Server
HP ProLiant BL465c G5
HP ProLiant BL685c G5
NOTE: Complete information on the HP ProLiant Servers can be found in the following QuickSpecs:
http://h18000.www1.hp.com/products/quickspecs/division/10455.html

All currently shipping processors for each control node are available. Select one or two processors (two or four processors with BL685c G5).

HP ProLiant BL465c G5
HP ProLiant BL685c G5
Configure servers as base for one or two compute pools. Total nodes cannot exceed 512 nodes.
NOTE: Complete information on the servers can be found in the following QuickSpecs:
http://h18000.www1.hp.com/products/quickspecs/division/10455.html

All currently shipping processors for each compute node are available. Select one or two processors (two or four processors with BL685c G5).

Numerous InfiniBand implementations are supported and can be selected based on the bandwidth requirements of the cluster. Typical clusters use 2:1 bandwidth since this satisfies the bandwidth requirements of most clusters. Some clusters select the higher 1:1 bandwidth. High Bandwidth (2:1) topology relies on eight uplinks for every sixteen node connections while the Full Bandwidth (1:1) utilizes twelve uplinks for every twelve node connections, avoiding bottlenecks.

Four switch sizes are offered for InfiniBand DDR cluster interconnect.

• The InfiniBand 288-port switch can be used as a standalone switch for configurations up to 288 nodes, in pair for configurations up to 384 nodes or as root switch in tree topologies with configurations up to 512 nodes. It must be equipped with two management boards, two or four fabric boards, up to twelve 24-port line boards and power supply modules in N+1 configuration.
• The InfiniBand 144-port switch can be used as a standalone switch for configurations up to 144 nodes.
• The InfiniBand 96-port switch can be used as a standalone switch for configurations up to 96 nodes
• The InfiniBand 24-port switch can be used as a standalone switch for configurations up to 24 nodes, in pair for (2:1) configurations up to 32 nodes, as root and node-level switch in tree topologies with configurations up to 96 nodes or as node-level switch in tree topologies with 288-port switch(es). In one switch or two switch solutions, the internally managed model must be selected. In multiple 24-port switch solutions, two internally managed models must be selected, others being externally managed models. When 288-port switch(es) are selected with some 24-port switches, all models must be externally managed models.
• The HP 4X DDR IB Switch Module for HP BladeSystem c-Class, a double wide switch module for the c-Class enclosure. The HP 4X DDR IB Switch Module for HP BladeSystem c-Class provides 24 InfiniBand 4X DDR ports with 20 Gbps port-to-port connectivity. The ports are arranged as 16 downlinks to connect up to 16 blade servers in the enclosure, and 8 uplinks to connect to the external InfiniBand switches to build an InfiniBand fabric.

All switches are Double Data Rate (DDR) models.

The HP ProCurve 2824 switch is only used as a standalone switch for small configurations up to 20 nodes (non-blocking).

The HP ProCurve 3500-24G switch can be used as a standalone switch for small configurations up to 24 nodes (non-blocking).

The HP ProCurve 2848 switch can be used as a standalone switch for configurations up to 40 nodes (non-blocking up to 40 ports), in pair (one) for configurations up to 64 nodes (8 links between switches), or as root switch or node level switch in tree topologies (8 uplinks/32 node ports or 4 uplinks/40 node ports).

The HP ProCurve 3500-48G switch can be used as a standalone switch for small configurations up to 48 nodes (non-blocking).

The HP ProCurve 5406zl switch can be used as a standalone switch for configurations up to 144 nodes (six 24-port 1000T modules) or as root switch with 4-port 10GigE modules in tree topologies for configurations up to 512 nodes (11 node level HP ProCurve 3500-48G switches attached by 2 10GigE uplinks each).

The HP ProCurve 5412zl switch can be used as a standalone switch for configurations up to 288 nodes or combined with another one for configurations up to 512 nodes (one 4-port 10GigE module in each switch to aggregate switches).

The HP ProCurve Switch 2626, 2650, 2848 and 2824 are used to build the internal administrative and console networks for the HP Cluster Platforms. These two networks are independent and organized in a tree topology with root switches and branch switches. The two root switches can be connected together based upon software or the end-user's requirements.

The console network is built only with either the HP ProCurve 2626 or the 2650 Fast Ethernet models, while the administrative network can use the HP ProCurve 2848 or 2824 models. Multiple uplinks between branch and root switches can be aggregated by trunking. The Gigabit administrative network is configured with 2 uplinks.

Each c-Class enclosure is equipped with an HP GbE2c Ethernet Blade Switch for c-Class BladeSystem (GbE2c).

• 48 10/100 ports
• 2 Gigabit link ports
• 10/100 auto-sensing per port automatically detects and sets the speed
• A comprehensive LED display with per-port indicators provides an at-a-glance view of status, activity, and speed
• Special 10000 rack-mounting kit to allow integration into rack for shipment

• 24 10/100 ports
• 2 Gigabit link ports
• 10/100 auto-sensing per port automatically detects and sets the speed
• A comprehensive LED display with per-port indicators provides an at-a-glance view of status, activity, and speed
• Special 10000 rack-mounting kit to allow integration into rack for shipment

• 44 10/100/1000 ports
• 4 dual-personality ports-each port can be used as either an RJ-45 10/100/1000 port or an open mini-GBIC slot
• A comprehensive LED display with per-port indicators provides an at-a-glance view of status, activity, and speed
• Special 10000 rack-mounting kit to allow integration into rack for shipment

• 20 10/100/1000 ports
• 4 dual-personality ports-each port can be used as either an RJ-45 10/100/1000 port or an open mini-GBIC slot

• Full set of industry standard Gigabit Ethernet Layer 2 features
• Sixteen internal downlinks and five uplinks and two internal cross-connects
• Wire speed switching on all sixteen 1Gb server ports
• Wire speed switching on all five 10/100/1000T uplink ports

A rackmount keyboard/monitor (RKM) is required with the HP Cluster Platforms. This 1U RKM with TFT display is attached to the control node that is designated as the cluster system administration node. The TFT display is mounted in the utility rack (UBB) and provides local user access to the cluster. In systems with utility nodes, an 8-port KVM switch is also available to connect the RKM to multiple utility nodes (up to 7) to allow for failover connectivity to other nodes. Systems with visualization nodes are also equipped with a KVM and necessary expander modules so all visualization nodes are connected to the RKM.

Rack 10000 series Cabinets
The increasing power of new high-performance processor technology requires increased cooling efficiency for rack-mounted servers. The 10000 Series Racks provide enhanced airflow for maximum cooling, allowing these racks to be fully loaded with servers using the latest processors.

HP Cluster Platform CBB, UBB, and UXR cabinets are equipped with three or four power distribution units while the IBB cabinet is equipped with two. These 1U units are installed front to back. They are all high voltage (200-240 VAC) models for all countries, equipped with appropriate power plugs:

• North America, Japan 24A (200-240 VAC): attached input cord NEMA L6-30P
• Optionally, North America, Japan 40A (200-240 VAC): attached input cord: Non-NEMA Locking CS8265C, 50A
• EMEA 32A (200-240 VAC): attached input cord IEC 309-32A

HP Cluster Platform 3000BL and 4000BL CBB are equipped with up to eight power distribution units. These 1U units are installed front to back. They are all high voltage (200-240 VAC) models for all countries, equipped with appropriate power plugs:

• North America, Japan 40A (200-240 VAC): attached input cord: Non-NEMA Locking CS8265C, 50A
• EMEA 40A (200-240 VAC): attached input cord IEC 309-60A

HP Cluster Platform VBB cabinets are equipped with one dual power monitoring power distribution unit that mounts on the cabinet rail. These PDUs are high voltage (200-240 VAC) models for all countries, equipped with appropriate power plugs:

• North America, Japan dual 40A (200-240 VAC): attached input cords non-NEMA, locking, CS8265, 50 A
• EMEA dual 32A (200-240 VAC): attached input cords IEC 60309, 32A 3-pin

HP Cluster Platform systems include one hardware documentation kit. Each hardware documentation kit contains both a CD and a hard copy of:

• HP Cluster Platform Installation and Operation Guide
• HP Cluster Platform Overview and Site Preparation Guide

The CD contains a soft copy of the same guides and Installation and User Guides of the various components integrated in the HP Cluster Platform system.

• The HP StorageWorks Modular Smart Array 1500 (MSA1500) is a 2 Gb Fibre Channel storage system designed for the entry-level to mid-range Storage Area Network (SAN). The MSA1500 is a scaleable, high performance storage system. With the addition of HP StorageWorks MSA30 or MSA20 disk enclosures, the maximum storage capacity is 16.8TB and 42TB respectively, when connected to the control node integrated in the UBB. Control node and utility nodes can be attached to a SAN, depending on the end-user's requirements and required system I/O bandwidth. Select one Fibre Channel host bus adapter for each designated node.

• Direct attached storage subsystems can be configured with HP StorageWorks MSA30 or MSA20 connected to Smart Array 6402 or 6404 integrated in the control node. The maximum storage capacity can be achieved when two Smart Array 6404 are configured in the control node. Then, five MSA30 or eight MSA20 can be directly attached and integrated in the UBB. This gives a maximum of 21TB and 48TB respectively.

• Direct attached storage subsystems can be configured with HP StorageWorks MSA60 shelves connected to Smart Array P800 controllers integrated in the control node. Up to eight MSA60 can be directly attached to one or two controllers and integrated in the UBB. This gives a maximum of 96 disk drives for a total of 72TB with 750GB SATA drives.

Linux

• Red Hat EL (RHEL) 4.0 and 5.0 (1 year and 3 year subscriptions)
  
  • Standalone server
  • HPC 8 pack for RH EL Basic Server 4.0 or 5.0 (1-2 processor nodes only with RH4.0)
• SUSE SLES 9 and SLES 10 (1 year and 3 year subscriptions)
  • Standalone server
  • HPC 8 packs: 1-8 processor nodes
• HPC 8 packs for Red Hat and SUSE support up to 8 compute nodes per 8 pack.
• XC System Software includes full Linux distribution, compatible with Red Hat EL 4.0

HPC 8 pack ordering information

1. Order one base kit for the 1st eight compute nodes: A standalone server subscription is recommended for control node (for SUSE SLES10, the base kit includes the subscription for the control node in addition to the first eight subscriptions for compute nodes. A standalone subscription is not necessary for the control node).
2. Order add-on kits to cover remaining compute nodes(1 kit per 8 nodes in case of SUSE SLES 10 and RH4.0, one kit per node in case of RH5.0)
3. 8 packs require purchase of HPS 10 incident care pack (one per cluster - E for electronic only)

HP XC System Software V3.2.1

The HP-developed XC System Software integrates leading open source cluster packages with the Linux operating system, HP-MPI, HP Scalable Visualization Array Software and Platform Computing's LSF. The XC System software provides single-system management for job and resource scheduling and simplified software installation, resulting in unprecedented levels of ease-of-use, productivity, and scalability.(NOTE: Users can add alternative MPI implementations and schedulers to an XC environment.)

On an XC System, all of the CPUs must be licensed. Standard software support includes 9x5 telephone support and rights to new versions for the specified period of time (one year or three years). 24x7 software support extends the telephone support window to round-the-clock access. See the HP XC System Software Version 3.2.1 QuickSpecs for more information and specific part numbers. A required Base license includes one set of media and documentation.

NOTE: Complete information on the XC System Software, including QuickSpecs, can be found at the following URL: http://www.hp.com/techservers/clusters/xc_clusters.html

The HP Cluster Management Utility is designed to manage a large number of compute nodes with a Single System View GUI (Graphics User Interface). HP CMU comes with a full Java GUI that can be tailored for your needs and for any number of nodes in the cluster. HP CMU is HP SIM (System Integration Manager) level 1 integration compliant. HP CMU provides a Command Line Interface (CLI) for day-to-day administration and cloning. Through its user friendly graphical interface HP CMU provides the cluster administrator with three main features:

• Administration: From HP CMU you can administrate your cluster (eg: halt, boot reboot, power off, broadcast commands from a single keyboard session)
• Cloning: For the first installation or for future updates, the cluster administrator has the capability to propagate a system configuration image to all or a part of the compute nodes in the cluster
• Monitoring: At a glance the manager can see the complete behavior of the cluster without to painfully analyze the performance of the individual compute nodes.
  In a CMU cluster configuration, all the compute nodes must be licensed. Every management node just needs a base license that includes HP CMU distribution and documentation. HP CMU software supports HP ProLiant Intel and AMD Opteron servers running Linux.
  

NOTE: Complete information on HP CMU software, including QuickSpecs, can be found at the following URL: http://www.hp.com/techservers/clusters/cmu_clusters.html

HP Insight Control Environment for Linux (ICE-Linux) provides comprehensive imaging, deployment, monitoring, and management for Linux-based HP ProLiant server platforms. Built on the industry-leading HP Systems Insight Manager (HP SIM) framework, this solution integrates open source technology with experience leveraged from HP's Linux products in HP ICLE and HP XC Clusters, while providing paths to the future with developing management technologies.

Why use HP Insight Control Environment for Linux?

• Integrated Linux management on industry-standard ProLiant servers:
  • Architected to manage expanding Linux multi-server ProLiant environments to meet the need for increased productivity, utilization, and control.
  • Leverage the best open source and HP management technologies using the established HP Systems Insight Manager (SIM) framework
  • Use established open source technologies (Nagios & others) for extensible health and performance monitoring
  • Enable workgroup and departmental clustering
  • Global support by HP

HP Insight Control Environment for Linux (ICE-Linux) brings the full expertise of HP's management investments from UNIX & Windows to the Linux environment, while providing flexibility and productivity to fulfill the variety of usual management tasks. HP ICE-Linux is a Linux-hosted solution on HP BladeSystem c-Class servers and rack servers (iLO and iLO2 capable). From bare-metal system discovery through health and performance monitoring, HP ICE-Linux enables maximum server control for productively managing Linux multi-server environments. Productivity, utilization, and control result from effectively managing Linux multi-server environments.

NOTE: Ordering information on HP ICE-Linux can be found in the QuickSpecs for the HP Insight Control Management Suites at the following URL: http://h18000.www1.hp.com/products/quickspecs/Division/Division.html#12949

For Cluster Platform solutions containing visualization nodes and running XC System Software, the HP Scalable Visualization Array Software provides the ability to identify and manage visualization resources and run visualization applications in the XC System environment.

NOTE: The HP Scalable Visualization Array Software is not available as a standalone software. It is included in the XC System software V3.2.1

HP Cluster Platform solutions warranties and service options are based on the offerings of the underlying components.

HP customer support provides onsite hardware break/fix support and remote, remedial software call-center support. Customer support offsite software services include level 1 and 2 support:

• Level 1 is defined as everyday user/system administration issues.
• Level 2 is defined as problems related to installation and configuration, along with other problems not solvable by following the vendor-supplied documentation.

The HP software support team will work in parallel with the appropriate vendor and development groups to address level 3 and 4 support elevations:

• Level 3 elevations typically require that patches and modifications be generated by the vendor to resolve deficiencies in the product.
• Level 4 elevations deal with enhancements in the functionality of the product that will typically be included in future releases.
  Optional customer services
  

HP has experience servicing systems, networks, storage devices, and software for HPC deployments and environments of all sizes. A wide range of purchasing options and solutions built from standards-based service modules ensure the service coverage and the level of expert assistance desired.

Optional services may be purchased at the time of the product purchase as HP Care Pack services, easy to order and easy to use service packages. Optional services may also be purchased at a later point of time as HP Contractual or Per-Event Services to extend and/or expand and/or complement the standard product warranty.

Hardware Support: Service capabilities include total lifecycle support, preventive and diagnostic services, access to a global service network, and support for an array of third-party peripherals.

Hardware Support Services offer a full range of high-quality remote and onsite hardware support options to meet specific response or repair-time objectives. These support options range from next-day or 4-hour onsite response to 6-hour call-to-repair time commitments. To accommodate specific hours of operations, coverage window options range from standard business hours (9x5), Monday-Friday 8am-5pm local time up to 24x7.

Software Support: Standard software support includes 9x5 telephone support and rights to new versions for the specified period of time (one year or three years). 24x7 software support extends the telephone support window to round-the-clock access. Standard software support includes customer access to technical resources during standard hours (see later), problem analysis, escalation management, and resolution. HP also provides unlimited access to an electronic facility that includes a knowledge database with known symptoms and solutions, software product descriptions, specifications, and technical literature.

Software support contracts can be obtained to meet the needs of customers, including remedial technical remote support along with migration and upgrade planning and a full suite of proactive deliverables.

NOTE: Additional information regarding worldwide limited warranty and technical support is available at: http://h18004.www1.hp.com/products/servers/platforms/warranty/index.html

C& I options include on-site systems software support and applications development and optimization, as well as clusters training. This is in addition to options for Factory and Field On-site installation (see above). Based on compute node counts.