1.2.7 Planning and evaluating your WebSphere HA solutions
We recommend that you use the following steps when planning and evaluating your end-to-end WebSphere HA solutions:
| 1.
| Analyze your requirements:
|
|
Do we need continuous operations? Most customers do not need continuous operations; therefore, upgrades of software and hardware can be performed offline.
|
|
What kind of availability do we need during your hours of operation?
|
|
What is the performance requirement during a failover? For example, the performance might be impacted because fewer nodes or servers serve client requests after a node/server failure. If this is not desired, provide ample extra resources.
|
| 2.
| Analyze the cost factors. How much will you lose when your system is unavailable, and how much can you invest in your system availability? If downtime is costly, you should invest an appropriate amount to improve your system availability.
|
| 3.
| Estimate the setup and administrative complexity. More complicated systems require skilled people and more configuration and administration effort.
|
| 4.
| Consider all the components in an end-to-end WebSphere system. Usually, the overall availability is dominated by the weakest point of the end-to-end WebSphere system chain. Consider the possibility of a failure in each component and its failover time.
|
| 5.
| Analyze failover time, which mainly includes fault-detection time and recovery time. For different failover mechanisms/techniques, the failover time is different.
|
| 6.
| Analyze the recovery point, where your processing resumes after a failover. It is directly related to the amount of work that is lost during a failover.
|
| 7.
| Understand the programming models. This concerns whether the failover is transparent to clients and whether one component's failover is transparent to other components in an end-to-end WebSphere system. Some services are able to perform failover transparently, because the failover is masked from the application. Others have options for adding application code that can retry when necessary.
|
| 8.
| Know that there is usually more than one solution to address a given failure. Some solutions might have special restrictions. Analyze the trade-offs between the different solutions.
|
|
