1.4.1 Continuous business capacity and performance
Ensuring that a system is able to handle a certain peak capacity and performance, as required by its business, requires eliminating single points of failure and building redundant capacity above and beyond the peak capacity. If part of the system goes out then the redundancy in the system ensures that 100% of the business capacity requirements can still be met by the system without any impact to the business functionality or performance.
Redundancy may need to be implemented at various levels, including:
- Hardware redundancy: In a clustered environment, such as the one employing WebSphere Application Server, this may refer to setting up horizontal cluster members, which are set up on different physical machines.
- Software redundancy: In a clustered environment, such as the one employing WebSphere Application Server, this may refer to setting up vertical cluster members, which are set up on same physical machine.
- Data redundancy: In a system that relies on a source of data, this refers to having more than one identical source of data. If one source of data goes down, the other sources of data continue to handle service requests. For example, DB2 HADR can provide one such solution (see Chapter 5, Database tier High Availability).
- Communications redundancy: This refers to the retry logic within an application whereby if the application is not successful in an operation that required communication outside of the application, then it retries the operation a certain number of times before signaling a failure. Retry logic may be implemented to identify faults in a certain hardware, software, or data source and to try the alternate, redundant, or standby hardware, software, or data source.
- Network redundancy: This refers to the hardware, such as bridges, access points, transmission cables, and so on (and the associated software), redundancy to ensure that the any network failures do not adversely impact a site's performance.
xxxx