User manual HP HIGH PERFORMANCE LX CLUSTER HIGH PERFORMANCE CLUSTERS LC SERIES - SETUP AND INSTALLATION GUIDE - TECHNICAL

[. . . ] high availability systems engineering july 2003 technical white paper HP High Performance Clusters LC Series setup and installation guide table of contents abstract introduction safety information HPC LC Series hardware solutions HPC LC Series solution components control node compute nodes HPC networks additional components 2 2 2 3 4 4 4 4 4 hp services and care packs receiving the HPC cluster installing switches physical planning positioning the racks rack cooling 5 5 6 6 7 8 cabling the configuration powering on the equipment factory system settings setting up the control node setting up the compute nodes summary for more information feedback 8 11 11 12 12 12 13 13 341524- 001 1 HP High Performance Clusters LC Series setup and installation guide abstract This setup and installation guide provides the customer with the information required to take the HP High Performance Clusters LC Series (HPC LC Series) product offering from delivery of order to a fully cabled system, ready for operating system and application installation. This document supplements the information found in the user guides for the servers and switches used in your HPC LC Series solution. introduction This guide provides the customer with instructions and the necessary reference information required for setting up and installing the HP High Performance Clusters LC Series solutions. The reader gains the knowledge of the various solution components and how to receive equipment, position racks, cable the configuration, power on the equipment, and begin the initial setup process. This paper provides information on each of the following topics: · · · · · · · · · · · HPC LC Series hardware solutions HPC LC Series solution components HP Services and Care Packs Receiving the HPC cluster Installing switches Physical Planning Cabling the configuration Powering on the equipment Default factory settings Setting up the control node Setting up the compute nodes This paper assumes that the reader has advanced technical skills and has extensive knowledge of the following topics and products: · · · High performance computing concepts Linux operating system installation knowledge and experience High performance computing system software installation experience safety information IMPORTANT SAFETY INFORMATION Before installation, read the Important Safety Information document included with the product. [. . . ] The 16-node solutions ship in a single 42U rack and require sufficient power for two (2) Power Distribution Units (PDUs). The 32-node solutions also ship in a single 42U rack and require sufficient power for three (3) PDUs. The 64-node solutions ship in two 42U racks and require sufficient power for six (6) PDUs. The 128-node solutions ship in four racks and require sufficient power for eleven (11) PDUs. When positioning the racks, be sure to place them in sequential order because the cable lengths are designed for that relative position. For example, figure 1 below illustrates the proper rack positioning of a 128-node cluster solution. figure 1. network overview of 10/100 Fast Ethernet and Gigabit Ethernet HPC solutions In figure 2 above, the DL380 server has one (1) iLO port and four (4) NIC ports. Both the iLO port and NIC1 port connect to the iLO switch (ISW). NIC2 is used to connect to an external network. NIC3 is used to connect to the cluster interconnect switch (CSW). In this configuration, the cluster interconnect switch can be either a 10/100 Fast Ethernet or Gigabit Ethernet switch. NIC4 is used to connect to the optional management switch (MSW). If the management switch is not part of the solution, then NIC4 will not be connected. The DL360 servers have one (1) iLO port and two (2) NIC ports. The iLO port connects to the iLO switch (ISW). NIC1 connects to the cluster interconnect switch (CSW). NIC2 connects to the optional management switch (MSW). If the management switch is not part of the solution, then NIC2 will not be connected. Note: The management switch (MSW) is an optional component for the 10/100 Fast Ethernet and Gigabit Ethernet solutions. Figure 2 above assumes you will be connecting your HPC LC Series solution to an external DHCP server. The DHCP server must be provided by the customer and be made available to the system before proceeding to use iLO. If you plan on using the control node and HPC cluster software to assign DHCP addresses, then you will need to cable the system accordingly. 9 HP High Performance Clusters LC Series setup and installation guide Figure 3 below provides a high level overview of the network cabling requirements for the Myrinet HPC cluster solutions only. iLO NIC1 iLO Switch (ISW) iLO Corporate Network NIC2 NIC3 NIC4 MYR Management Switch (MSW) NIC1 NIC2 Interconnect Switch (CSW - Myrinet) MYR DL380 Server DL360 Servers CAT5E (internal rack cables) CAT5E (external rack cables) Fibre optic cables figure 3. [. . . ] Note: During the installation process you may be required to assign each NIC a specific function within the HPC cluster. setting up the compute nodes The ProLiant DL360 servers ­ compute nodes can be powered on in any order. The compute nodes are not connected to a keyboard, monitor, or public LAN. Therefore, you must use the integrated Lights-Out (iLO) feature of the ProLiant servers in order to remotely manage each compute node. Each ProLiant machine has the iLO Advanced license installed. [. . . ]