HP WebQoS Administration Guide

Understanding HP WebQoS
The Role and Benefits of HP WebQoS
Chapter 1 15
This is controlled by the WebQoS policies. When new sessions are
rejected, redirected, or deferred, existing sessions continue
uninterrupted. User request prioritization and session management are
the primary tools used by WebQoS to provide capacity protection. The
network manager is responsible for updating the timers that define a
session, and for establishing the policies that control if the session is
admitted.
Capacity Protection
Capacity Protection prevents system overload with prioritization and
admission control. This minimizes the impact of unexpected surges in
demand while maximizing the volume of completed transactions.
Capacity Protection makes sure that performance levels for active
customers and their currently running transactions are not
compromised. New user requests are not admitted to a site unless their
transactions can be completed quickly.
Capacity is assessed by service level objectives and thresholds.
Protection is provided by corrective actions.
SLOs, Thresholds, and Corrective Actions
To ensure that your business policies are reflected in your Web
applications, they need to be translated into service level objectives
(SLOs) and thresholds. SLOs are business-oriented policies and
thresholds are operations-oriented policies. When an SLO or threshold is
violated, a configured corrective action is performed.
SLOs WebQoS has the following types of business-oriented SLOs:
• Capacity - Maximum number of concurrent sessions supportable with
acceptable performance. Typically coupled with a response time SLO
or performance threshold.
• Response times - A key component of customer satisfaction.
Measured from the time a request enters the server to the time a
response is sent back.
• External measurement - Measurements collected from external
sources can be used to determine corrective actions.
Thresholds WebQoS uses an operations-oriented load threshold to
keep your system operating properly during times of peak load. System
load is assessed as a function of CPU (central processing unit) usage.