Table 3-1. Service procedure tools

Utility (see note)

Runs on

Description

 

 

 

fault_service_Worm_RTG

All nodes

Monitors the switch for faults. It restarts the switch if a

 

 

fault is detected.

 

 

 

fs_monitor

All nodes

Monitors the adapter for interrupts that have not been

 

 

serviced. This is required because of a hardware

 

 

oversight in MSMU, in which there is a window in time

 

 

that causes it to hold interrupts and then forget to

 

 

introduce them.

 

 

 

rc.switch

All nodes

Starts the daemons and IP configuration tools.

 

 

 

Estart

Primary or

Tunes the switch and puts it into run phase. Also, kicks

 

Control Node

off the route table generator and distributes the routes to

 

 

the nodes, by placing them in the etc/SP directory on

 

 

the control node.

 

 

 

Estart_sw

Primary or

This is the primary-node command to start the RS/6000

 

Control Node

SP High Performance Switch. Estart_sw is called by

 

 

Estart. It is useful to note that the time-out variable

 

 

(LIMIT) for the Estart process is located in this file. This

 

 

may have to adjusted for systems that take longer to

 

 

tune than in other systems.

 

 

 

Eprimary

All nodes

Sets a node as the primary node.

 

 

 

Eprotocol

All nodes

Stores switch protocol information in the System Data

 

 

Repository for a node or a series of nodes. Also

 

 

retrieves a topology files out of the SDR.

 

 

 

Eclock

Primary or

Controls the clock source for each switch board within

 

Control Node

an SP cluster.

 

 

 

ifconfig

All nodes

This sets up the IP interface to the switch.

 

 

 

Note: Unless otherwise noted, the directory for these utilities is /usr/lpp/ssp/css.

 

 

 

Table 3-2. Setup output files

 

 

 

 

 

File (see note)

Location

Description

 

 

 

rc.switch.log

All nodes

Logs all information on the last run of rc.switch. It will

 

 

indicate if this is a primary or secondary node, the

 

 

associated switch chip information, and IP address

 

 

information.

 

 

 

rc.switch.log.previous

All nodes

A copy the previous run of rc.switch.

 

 

 

expected.top.no_comments

All nodes

Used by rc.switch to more easily parse out chip

 

 

connection information.

 

 

 

fs_daemon_print.file

All nodes

A log of the daemon.

 

 

 

fs_monitor.log

All nodes

A log of the monitor daemon.

 

 

 

css.snap

All nodes

Log files created by the switch support code.

 

 

 

Note: The directory for these utilities is /var/adm/SPlogs/css.

Table 3-3. Tuning output files

File (see note)

Location

Description

 

 

 

daemon.stdout

Primary

Keeps a detailed account of the tuning process initiated by the

 

 

Estart command. It includes data from every tuning operation

 

 

since the current daemon on the primary node was initiated.

 

 

 

daemon.results

Primary

A record of how many nodes were initialized.

 

 

 

3-2RS/6000 SP: SP Switch Service Guide

Page 72
Image 72
IBM RS/6000 SP Utility see note Runs on Description FaultserviceWormRTG, Fsmonitor, Rc.switch, Estartsw, Eprimary

RS/6000 SP specifications

The IBM RS/6000 SP (Scalable Power) is a high-performance computing system that was developed in the early 1990s and has since become a significant player in the realm of supercomputing and enterprise solutions. Designed primarily for scientific, engineering, and complex data analysis tasks, the RS/6000 SP provides powerful processing capabilities, enhanced scalability, and features to support demanding applications.

One of the key characteristics of the RS/6000 SP is its modular architecture, which allows for the addition of multiple nodes. Each node is based on IBM's Power architecture, leveraging the RISC (Reduced Instruction Set Computing) design to achieve high throughput and efficiency. The system can support configurations ranging from a few nodes to hundreds, enabling organizations to scale their computational power as needed based on workload requirements.

The RS/6000 SP employs advanced interconnect technologies, most notably the SP Switch, which ensures high-speed communication between nodes. This interconnect allows for seamless data transfer, ensuring that the system can handle large datasets and complex computations without bottlenecks. The SP's ability to deliver concurrent processing capabilities makes it ideal for parallel computing tasks, including simulations, modeling, and large-scale computations.

In terms of software, the RS/6000 SP was equipped with AIX, IBM's version of the UNIX operating system, which provides a stable and secure environment for enterprise applications. Furthermore, the system supports a vast array of programming languages and tools, including Fortran, C, and C++, along with libraries and frameworks designed for high-performance computing. This compatibility allows developers to optimize their applications to fully utilize the underlying hardware.

Another notable feature of the RS/6000 SP is its extensive support for various workloads, including database management, multi-user applications, and graphical processing. The system was equipped with high-performance graphics capabilities, making it suitable for tasks that require intensive visualization, such as computer-aided design (CAD) and scientific visualization.

Overall, the IBM RS/6000 SP has solidified its role in the high-performance computing landscape, combining advanced technologies, robust scalability, and flexibility to meet the unique demands of research and enterprise environments. Its legacy continues to influence modern computing architectures, embodying IBM’s commitment to innovation and performance in the pursuit of complex problem-solving capabilities.