29.01.2013 Views

WebSphere Application Server V7.0: Concepts ... - IBM Redbooks

WebSphere Application Server V7.0: Concepts ... - IBM Redbooks

WebSphere Application Server V7.0: Concepts ... - IBM Redbooks

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

4.3 Sizing the infrastructure<br />

After determining the initial application and infrastructure design and scalability<br />

techniques, you need to determine the system resources required for the project<br />

to keep the service level agreements (SLAs). Whenever making sizing decisions<br />

ensure that all design decisions are still honored.<br />

<strong>Application</strong> design will evolve over time, and sizing is usually done in the early<br />

stages of design. However, when sizing, it is important that you have a static<br />

version of the application design with which to work. The better view you have of<br />

the application design, the better your sizing estimate will be.<br />

You have to consider which hardware platforms you want to use. This decision<br />

depends on the following factors:<br />

► Scaling capabilities of the platform<br />

► Platforms <strong>WebSphere</strong> <strong>Application</strong> <strong>Server</strong> supports<br />

► Performance, security, and high availability requirements of your environment.<br />

Next, determine your scaling approach:<br />

► Scaling up—vertical scaling<br />

Scaling up is done inside a single system, typically on a multi-processor<br />

machine. Even if this approach might be sufficient from the performance<br />

perspective, the underlying hardware is the single points of failure.<br />

► Scaling out—horizontal scaling<br />

Scaling out means increasing the number of machines. Scaling out is typically<br />

done to improve high availability by limiting single points of failure as well as<br />

for scalability reasons when hardware resources are the limiting factor. You<br />

have to support and maintain multiple machines. If you decide to use the<br />

scale out approach keep in mind that your design must be capable of<br />

processing the workload even if one system fails.<br />

Sizing estimates are based solely on your input, which means the more accurate<br />

the input, the better the results. Sizing work assumes an average standard of<br />

application performance behavior and an average response time is assumed for<br />

each transaction. Calculations based on this are performed to determine the<br />

estimated number of machines and processors your application will require. If<br />

your enterprise has a user experience team, they might have documented<br />

standards for a typical response time that your new project is required to meet.<br />

If you need a more accurate estimation of your hardware requirements and you<br />

already have your application, consider using one of the benchmarking services<br />

discussed in 4.4, “Benchmarking” on page 107.<br />

106 <strong>WebSphere</strong> <strong>Application</strong> <strong>Server</strong> <strong>V7.0</strong>: <strong>Concepts</strong>, Planning, and Design

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

Saved successfully!

Ooh no, something went wrong!