 |
» |
|
|
|
Single-User
versus Multiuser | |
For the purposes of this document, we’ll be distinguishing
between two ways for people to use a given system: as a single-user workstation,
usually on someone’s desk and used mainly or exclusively
by that person; as a multiuser system, often
kept in a computer room, with which individual users communicate
by means of a terminal, or terminal-emulator on a desktop system
connected by a LAN or modem. The power of stand-alone systems to handle more and more users
(as well as many other network functions) has grown dramatically. For this reason, if you plan to set up a standalone machine
as a multiuser system, refer to the information regarding On-Line Addition
and Replacement in the Configuring HP-UX for Peripherals book.
This material can help you to plan your system set-up so that in
the event of certain hardware failure, you can replace the hardware
with little impact to the users.
Server
versus Client | |
Broadly speaking, a server supplies
some kind of computing resource (applications, files, compute cycles,
printing and spooling...) and a client uses
that resource. In this document, we’ll use the terms server and client most
commonly, although not exclusively, in the context of NFS (Networked
File System) services, and we’ll make that context clear
wherever necessary by using the terms NFS server and NFS
client. Under NFS, and in most other contexts, the same system can
function as both a server and a client. For example, a system may
import one file system (NFS-mounting it from another system’s
disks) while exporting another (allowing other systems to NFS-mount
the file system from its own disks). As an importer of one or more
file systems, the system acts as an NFS client; as an exporter,
it acts as a an NFS server. Partitioned
Systems (The Partitioning Continuum) | |
HP-UX 11i provides many ways to isolate or combine system
resources (for example CPUs, memory, and I/O cards). HP refers to
the collection of system administration solutions that provides
these capabilities as the Partitioning Continuum. In addition to the traditional operating mode of one HP-UX
operating system per computer, multiple instances of HP-UX can run
on a single computer (using partitioning), and multiple computers
can be combined to host a single instance of HP-UX (using clustering).
Depending on the versions of HP-UX you are using and the hardware
you are running them on, you can use any of the following technologies
(or combinations of them) to maximize the efficiency and flexibility
of your HP-UX-based equipment: - iCOD
Instant Capacity on Demand allows you to have processors
(CPUs) in your system that you have not yet purchased. These processors
sit idle until you activate them using special iCOD commands. For
more information on iCOD, go to the “On Demand Solutions (ODS)” section
of http://docs.hp.com/hpux/netsys. - nPartitions
Some PA-RISC and Itanium-based Enterprise Server systems
(for example, Superdomes) have processors, memory, and I/O interface
connections mounted on cell boards. These
systems usually contain multiple cell boards. | | | |  | NOTE: I/O cards
plug into I/O chassis. The chassis are connected to (associated
with) the cell boards by cables or other internal connections. | | | | |
Cell boards can be grouped in specific ways such that the
resources of each group are electrically isolated from the resources
of other groups in the system. These groups, called partitions (sometimes
referred to as hard partitions, or physical partitions), separate
the resources of the computer into self-contained units, protected
from each other in such a way that hardware or software crashes
in one partition are unable to affect the operations of neighboring
partitions. Each partition is capable of supporting its own operating
system[1]. The term nPartitions derives
from algebra, where the “n” refers to a variable number,
indicating that you can group (and regroup) the cell boards in your
system in different ways to create varying numbers and sizes of partitions
(to best suit your needs). The smallest unit of construction in an nPartition is a cell
board (that is, you cannot use hardware partitioning to subdivide
the resources of a cell board, assigning them to more than one partition).
To do that, use Virtual Partitions (see vPars
below). nPartitions can be configured and managed using the ParMgr
tool (see parmgr(1M), or by various command line
tools. The following manpages describe nPartition commands (for
use from the command line): Table 1-1 nPartitions Manpages: Extensive information on using and configuring nPartitions
is located in the HP System Partitions Guide . - PPU
Closely related to ICOD, the Pay-Per-Use
technology allows you to pay for your HP-UX-based equipment based
on your actual usage rather than by purchasing hardware directly.
Detailed information on ICOD and PPU can be found in Instant
Capacity on Demand (iCOD) User’s Guide for Version B.05.00 (see
the “On Demand Solutions (IDS)” section of http://docs.hp.com/hpux/netsys). - PRM
Process Resource Manager is a resource management tool
used to control the amount of resources that processes use during
peak system load (at 100% CPU, 100% Memory, or 100% disk bandwidth
utilization). PRM can guarantee a minimum allocation of system resources available
to a group of processes through the use of PRM groups. Detailed
information on PRM can be found in HP Process Resource
Manager User’s Guide and HP-UX Workload
Manager User’s Guide - PSETS
Processor Sets allow a multi-processor system to
be partitioned into two or more groups of processors (CPUs) within
a given instance of HP-UX, so that CPU resources for
selected applications or users can be isolated from those of other
applications or users. - vPars
If you have a multi-processor system (whether
or not it supports nPartitioning), or if you want to
further subdivide the resources of an nPartition in a machine that
supports nPartitioning (See “nPartitions”),
you can use virtual partitions. Virtual partitions provide more flexibility than nPartitions
and they provide the same protections against software/operating
system crashes that nPartitions provide; however, a crash
due to a hardware failure will bring down all operating systems in
all virtual partitions within the machine or nPartition where the
failure occurred. vPARS can be configured and managed from the Virtual Partition
Manager, or using the command line. Extensive information on installing
and configuring vPARS is located in Installing and Managing
HP-UX Virtual Partitions (vPars). The following manpages describe vPARS commands (for use from
the command line): Table 1-2 Vpars Manpages: Not all HP-UX-based machines support virtual partitions. For
detailed information on which machines and HP-UX releases support
vPars, see Installing and Managing HP-UX Virtual Partitions
(vPars). - WLM
WLM expands on the features of PRM by providing
a more dynamic way to allocate resources. WLM automatically configures
PRM based on criteria you define (known as Service Level Objectives
-- SLOs), and by regularly monitoring resource availability. By specifying
the level of service you expect from your computer and applications
you have given WLM its goals. WLM works with PRM and the HP-UX schedulers
to achieve and maintain these service-level goals.Detailed information
on WLM can be found in HP-UX Workload Manager User’s
Guide
Hardware | |
The systems discussed in this document are mainly: HP 9000 Enterprise Servers
Operating
Systems | |
This document is for administrators of HP-UX systems, and
the workgroups we describe are predominantly made up of such systems, with
some PCs running Microsoft Windows or Linux operating systems. |
Printable version
