Extras din proiect
Abstract
Nowadays, when businesses have a considerable and growing World Wide Web presence, it is increasingly vital for them to have quantitative measurement tools of the response times of their web services.
Current approaches include active probing from geographically distributed monitors, instrumenting HTML Web pages with JavaScript, online monitoring, offline analysis of packet traces, and instrumenting Web servers to measure application-level performance or per connection performance. Some of these ones fall short, in one area or another.
This paper presents a comparison of performance and flexibility between such measurement utilities which can provide mechanisms for filtering and analyzing packets, whether they are implemented in kernel space or user space.
1 Introduction
“Some users and applications drive the revenue of the business. If the system is slow, customers go elsewhere, and transactions or sales are lost forever.” - P. Sevcik, Business Communications Review.
Response time is a key indicator of end user satisfaction in using web services. Customers seeking quality online services have choices, and will simply take their business elsewhere when response times exceed acceptable thresholds. As dependence on the World Wide Web continues to grow, it is increasingly important for businesses to understand and manage the response times experienced by their clients.
The ability of a Web hosting center to move CPU cycles, machines, bandwidth and storage from a hosted Web site that is meeting its latency goal to one that is not, is a key requirement for an automated management system. Such allocation decisions must be based on accurate measurements. Over allocating resources to one hosted Web site results in an overcharge to that customer and a reduction in the available physical resources left to meet the needs of the others. Under-allocation results in poor response time and unsatisfied Web site users. The ability to base these allocation decisions on a measure that is relevant to both the Web site owner and the Web site end user is a competitive advantage.
The term itself, “response time”, has become diluted, meaning a variety of different metrics to different people: system administrators, database administrators, network administrators. From the remote client perspective, there is only one measure of response time that matters: how long it takes to download a Web page along with all its embedded objects.
Preview document
Conținut arhivă zip
- Estimation Tools for Determining the Response Times of Web Services.doc