About This Guide

This guide introduces the administration and programming features of in the Silicon Graphics, Inc.® Array 3.0 software product. Array 3.0 software supports Silicon Graphics, Inc. Array systems. Array systems are affordable, scalable systems that are used both by commercial users who require ultra-reliable database and file servers, and by scientific users who require high performance.

An Array system, as supported by Array 3.0, is a cluster of Silicon Graphics, Inc. systems connected by a network (for details, see “Array Nodes”). Array 3.0 software can also run on individual multiprocessors.

Specific hardware packages that have been sold as Array systems include POWER CHALLENGEarray, CHALLENGE® DataArray, and other names. In this book, all systems on which Array 3.0 can run are collectively called Array systems.

Array systems are marketed to different audiences and used for different kinds of work, but the architectural concepts are the same, and so are the software tools used by system administrators and by software developers.


This guide is written for three groups of people: users who want to run software on an Array; administrators who need to configure an Array; and software developers who want to write programs for an Array.

An Array system is composed of many layers of hardware and software features, and each feature is documented separately. The main purpose of this guide is to help you orient yourself amid the profusion of printed and online documentation that is available.

If you are a system administrator, this guide introduces you to the information and software tools you use to configure an Array system, tune it, and keep it running.

If you are a software developer, this guide introduces you to the development tools and libraries you use to write programs that do high-performance parallel computation on an Array system.

Structure of This Guide

This guide contains the following chapters and appendices:

Conventions Used in This Guide

This manual uses the conventions and symbols shown in Table i.

Table 1. Typographical Conventions

Type of Information

How It Appears in the Text

Filenames and pathnames

This structure is declared in /usr/include/sys/time.h.

IRIX command names and options used in normal text

Update these variables with systune; then build a new kernel with autoconfig -vf.

Names of program variables, structures, and data types

Global variable mainSema points to an IRIX semaphore, which has type usema_t.

Names of IRIX kernel functions, library functions, and functions in example code

Use mmap() to map an object into the address space, and munmap() to remove it.

User input within a multiline example

tokyo% ainfo ash

Names of IRIX reference (“man”) pages. You can click on any of these to display the page. To dismiss the page, use alt-F12.

See the plock(2) reference page.

URLs on the World-Wide Web. You can click on any of these to launch Netscape.

Visit http://www.sgi.com .

Using Clickable References

Names of reference pages in red can be clicked to open a shell window displaying the output of the man command for that page (see the man(1) reference page for practice).

Note: If the reference page has not been installed on your system, the window will only display an error message.

Tip: You can quickly dismiss a shell window with the alt-F12 keystroke in that window.

URLs that appear in red can also be clicked. The URL is sent to Netscape, which displays the cited page.

Note: Netscape can take many seconds to start up if it is not active, and several seconds to appear if it is iconized when you click the reference. Be patient when clicking URLs.