|
7.1.1 Availability considerations
An IBM WebSphere Application Server Network Deployment V6.1 environment will likely include one or more of each of the following processes as part of the WebSphere specific infrastructure:
| 
|
| WebSphere Node Agent processes
|
|
|
|
| WebSphere Application Server processes
|
|
|
|
| Database server processes
|
Making an entire end-to-end system highly available will include all of these and more, including any back-end computers the system relies on, any networking components such as routers, switches and firewalls, any distributed storage components such as SAN, NAS, or Networked File Systems, any remote security solutions with which the system must integrate, and so on.
This chapter focuses entirely on the WebSphere processes themselves, that is, the Deployment Manager, the Node Agent, and the Application Server instances. Keep in mind that this is just one part of the system, and the whole system must be made highly available and performant in the same way in order to be of any value.
When using a clustered IBM WebSphere Application Server Network Deployment V6 environment, WebSphere process failures usually do not contribute much to the total number of client request failures, because the WebSphere process failover is instantaneous. However, a more serious failure (such as a physical problem with the machine on which an LPAR is running) can be mitigated by using HACMP.
Moreover, the HACMP configuration, when running on AIX/POWER 5, can be configured to automatically request and assign the resources required to the backup LPAR, meaning that these backup resources are not utilized until required. In a traditional non-LPAR environment, these resources would typically be machines acquired to stand idle, in case of failure, which is far less cost effective than using AIX/System p LPARs.
|