Certification Study Guide: IBM Tivoli Access Manager for e-business 6.0

62 Certification Study Guide: IBM Tivoli Access Manager for e-business 6.0

2.6.2 Sizing and availability

Availability is the major concern that a failing part of the infrastructure will cause

the overall solution to languish. This eventually leads to unsatisfied customers

and decreasing business success. Adding replicas of crucial servers increases

your site’s availability by avoiding single point of failure. All Access Manager

components can be replicated with the exception of the Policy Server, since

there can be only one Policy Server per secure domain. However, there can be a

second Policy Server in standby to provide manual fail-over capabilities as a first

aid response. If you want to assure 24x7 availability of your Access Manager

Policy Server you could implement a high-availability cluster solution such as

HACMP for AIX.

Before configuring a standby Policy Server, the files that it needs to operate must

be made available. To avoid synchronization problems, it is best to locate these

files on a shared file system.

In general, the most effective way to have a redundant Policy Server is to

configure an

original and standby Policy Server in an HACMP (or similar)

environment. This handles routing IP traffic to the active instance and can handle

(via scripting) the starting and stopping of the Policy Servers so that only one is

active at any time. IBM supports automatic failover only on the AIX platform.

Figure 2-12 shows a possible configuration that uses a network load-balancer to

direct SSL traffic to the active Policy Server. If it is not possible for the load

balancer to monitor the Policy Servers, then manual intervention (or custom

scripting) will have to be used to monitor the Policy Servers and switch to the

backup on failure.