Highly Available WebSphere Business Integration Solutions

Chapter 2. Overview of high availability concepts 15

availability on the order of, for example, 24x7x365 availability, the duration of a

planned outage can be minimized by applying techniques such as rolling

upgrades for software and hot replacement of hardware. Additionally, these

events can be scheduled during periods of lighter system load.

Reasons for planned outages include:

򐂰 The installation of software upgrades or patches

򐂰 Periodically scheduled backups

򐂰 Hardware repairs or expansions

򐂰 The physical movement of applications of hardware

Unplanned outages are the more urgent concern as they may occur at a time

when the system is required to be operational. The major contributors to

unplanned downtime are:

򐂰 Hardware failure

򐂰 Software failure

򐂰 Human error

򐂰 Environmental conditions: climate control, power failure, power spike

򐂰 Natural disasters: fire, flood, earthquake

2.4 The focus of this redbook

In this book, the primary focus is on solutions to help manage hardware and

software failure. Examples of the types of failures the given solution addresses

are:

򐂰 Network failures

򐂰 Disk failures

򐂰 Processor failures

򐂰 Application software defects

򐂰 System software defects

There are several different ways to accomplish the various levels of system

availability. These include the combination of software, hardware, and

operational procedures that work together.

2.5 Clustering for high availability

Clustering solutions that provide failover support are the prevalent mechanism

used to achieve high levels of availability.

16 Highly Available WebSphere Business Integration Solutions

IBM High Availability Cluster Multi-Processing for AIX (HACMP) is the platform

specific clustering solution provided by IBM for pSeries servers running AIX. HA

clustering solutions are provided by many vendors; among these solutions are:

Sun Cluster, Microsoft® Cluster Server (MSCS), VERITAS Cluster Server

(VCS), HP Serviceguard, and many others.

These clustering solutions are transparent to the applications that run under their

control. Applications do not have to be modified before they can be deployed

within an HA cluster.

Commercial HA clustering solutions allow for a wide range of cluster

configurations. The simplest cluster configuration is comprised of a pair of

servers, and an external SCSI or Fibre Channel storage subsystem.

The following terms are generally used when discussing cluster technologies.

Cluster

A cluster is a grouping of two of more interconnected computers that are viewed

and used as a single computing resource.

Node

A node is an individual computer system (including its hardware resources such

as local disks, processors, memory, etc.) which runs both the operating system

and the corresponding clustering software.

Nodes can be uniprocessors, or as in the case of a large scalable system,

symmetric-multi processors (SMP). An SMP node looks like a uniprocessor node

to most of the clustering technologies. Each node in the cluster can vary in the

number of processors. The number of nodes that can participate in a single

cluster varies depending on the clustering technology being implemented.

Resource groups

Resource groups enable the combination of related resources into a logical

entity. Most of the available systems clustering technologies employ the concept

of moving the system resources that are contained within a resource group.

Resource groups facilitate moving critical workload from one node to another,

and the continuation of critical workload processing on a different cluster node.

This concept will become more prevalent as we discuss system and application

availability configurations, especially in the case of MQ and Message Broker.