Highly Available WebSphere Business Integration Solutions

46 Highly Available WebSphere Business Integration Solutions

3.7 Capacity planning

In an HA cluster environment where under normal operations each node only

runs one application, the nodes will be configured for mutual takeover.

Therefore, the IT professional must consider that in a failover scenario, each

node should be capable of supporting both applications. To gather the hardware

requirements, one must consult the standard system installation guides of each

application. However, there are certain requirements that are specific to an HA

environment:

򐂰 The requirement is that you have at least two nodes. For example, you could

have two IBM pSeries 670s, or you could have a single IBM pSeries 690. The

reason that a single p690 will work is that it can be configured into multiple

partitions and each partition can be a node. The pSeries 690 delivers true

logical partitioning (LPAR). Each system can be divided into as many as 16

virtual servers, each with its own set of system resources such as processors,

memory, and I/O.

򐂰 A shared disk system such as IBM’s FAStT storage servers or IBM’s

7131-405 SSA disk subsystems. Application data and log files are written to

the shared drive frequently during normal operations. So, it is important to

ensure that there is a fast connection to the shared storage from both nodes

such as a fibre switch. And to increase availability, one may want to provide a

redundant fibre connection from each node to the shared storage.

򐂰 A device over which the cluster can exchange heartbeats. This can be a

public TCP/IP network, a private point-to-point network, or a shared disk.

򐂰 A public IP network to communicate with the outside world. Again, the IT

professional must be aware that in the case of a failover, one node must be

able to accommodate the work of both nodes. If you use IPAT through IP

aliases, you have to provide as many different subnets as you have resource

groups. If you use IPAT through IP replacement, every node needs to have

as many NICs as the total of all resource groups in the cluster.

Refer to the High Availability Cluster Multi-Processing for AIX Planning and

Installation Guide Version 5.1, SC23-4861 for details of how to select appropriate

disk technologies and networks.

3.8 Maintenance planning

Maintaining an HA system means maintaining hardware and software changes

without causing any unnecessary system outages. Any changes to a highly

available production system should be done in the period when there is the

lightest load on the system. Although WebSphere Business Integration can

handle unintended shut downs while events are in progress, any maintenance