25.02.2013 Views

TCP/IP Tutorial and Technical Overview - IBM Redbooks

TCP/IP Tutorial and Technical Overview - IBM Redbooks

TCP/IP Tutorial and Technical Overview - IBM Redbooks

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

24.1 Availability<br />

24.2 Scalability<br />

Application instances, network interfaces, <strong>and</strong> machines can fail (planned for<br />

maintenance or unplanned due to application or system error). In these cases,<br />

users must not lose their service. Recovering from application instance failure is<br />

fairly straightforward in that the application is simply restarted. Network interface<br />

failures can also be tolerated by making use of a virtual <strong>IP</strong> address, which is not<br />

tied to any particular physical interface <strong>and</strong> thus will never fail.<br />

A virtual <strong>IP</strong> address can be given to a device that has one or more network<br />

interfaces. This allows the users’ machines to pick up a specific <strong>IP</strong> associated<br />

with a specific machine or device. However, the <strong>IP</strong> address given is not tied to<br />

the physical <strong>IP</strong> address of the device’s interfaces. Therefore, if one of the<br />

interfaces fails, the users are unaware of the failure.<br />

Machine failure, however, is a bit more complex. Users must be able to<br />

immediately reconnect to the service without knowing that they now are using an<br />

alternate image of the application on another system. Users also must not be<br />

aware that the path to the other system has been automatically changed. The<br />

use of virtualization can be very advantageous with regards to increasing the<br />

availability of a system. Virtualization is discussed further later.<br />

Scalability means to provide a solution for a growing business that requires<br />

additional system capacity. When workload capacity becomes smaller due to<br />

many more new connection requests from clients or business partners, a<br />

nondisruptive growth of the current system environment must be made available.<br />

In a traditional single system environment (no clustered systems), a<br />

nondisruptive upgrade of systems is relatively limited. In order to raise capacity,<br />

these systems have to be taken down to install new features. Therefore, they are<br />

not available for a certain time.<br />

The implementation of clustered systems is a better approach (discussed in<br />

more depth later). Adding a new system to the cluster running equal applications<br />

instances does not impact the other systems in the cluster. This solution adds<br />

seamless capacity for a growing business. Compared to traditional systems, the<br />

user is not bound to a given system in a clustered server environment.<br />

Therefore, the management of user connections to servers is more flexible.<br />

When a new system comes online, new connections are directed to that machine<br />

taking over a new workload.<br />

Chapter 24. Availability, scalability, <strong>and</strong> load balancing 909

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!