IBM RS/6000 SP, Руководство пользователя

Страница: 29 / 480

Chapter 1. Introduction
5
Each processor node contains its own copy of the standard AIX operating
system and other standard RS/6000 system software, providing access to
thousands of available AIX applications. A set of new software products
designed specifically for the SP (PSSP with Virtual Shared Disks (VSD),
Recoverable Virtual Shared Disks (RVSD), Concurrent Virtual Shared Disks
(CVSD), General Parallel File Systems (GPFS), and Parallel Environment (PE))
allows the parallel capabilities of the SP system to be effectively exploited.
The SP processor nodes, on SP Switch or SP Switch-less configurations, can be
partitioned into pools of nodes. For example, four nodes can work as a Lotus
Notes server, while eight others process a parallel database. The SP system
partitioning is designed to divide the system into non-overlapping sets of nodes
in order to make the system more efficient and more tailored to your needs.
On an SP/CES processor node, the resources may also be partitioned using the
AIX WorkLoad Manager (WLM) facilities. For example a part of the processor
and physical memory resources can be dedicated for a specific class of jobs - for
example, interactive applications, and another part can be dedicated for other
specific class of jobs - for example, batch applications. WLM is designed to give
system administrators greater control over how the scheduler and Virtual
Memory Manager (VMM) allocate CPU and physical memory resources to
processes. This can be used to prevent different classes of jobs from interfering
with each other and to allocate resources based on the requirements of different
groups of users.
The RS/6000 SP system takes the advantages of the industrial leading RS/6000
workstation technology, the latest RS/6000 processors are repackaged for use
as SP nodes, and the SP nodes are interconnected by a high-performance,
multistage, packet-switched network for inter processor communication in order
to perform the parallel functions.
The SP/CES system optimizes high availability through built-in redundancy,
subsystem recovery, component error checking and correction, RAID5, external
and internal disk mirroring, and hardware and software monitoring. High
availability clusters of up to 32 SP nodes are supported by one of the industry’s
leading software products for critical application backup and availability, High
Availability Cluster Multi-Processing (HACMP) for AIX. If an error such as a node
failure occurs, the system can execute a recovery script that transfers the work to
another node and prevents the application from going down. RS/6000 cluster
technology, implemented also on the SP/CES systems, provides a collection of
services (topology, group, event management) that define hardware and
software resources, node relationships and coordinated actions to manage
groups of nodes.