dc.description.abstract | As enterprises world-wide race to embrace real 吃time management to improve
productivity, customer services, and flexibility. Many resources have been invested in enterprise systems (ESs). All modern ESs adopt a n-tier client-server architecture that includes several application servers to host users and applications. As in any other multi-server environment, the load distributions and user
distributions in particular, become a critical issue in tuning system performance.
In ESs, each application is evoked by a user who logs on an application server
and stays connected to the server for an entire working session, which can last
for days. Therefore, admitting a user into an application server affects not only
current but also future performance of the server.
Distributions in application servers and web servers are different in granularity. In the former scenario, a user represented by a set of transactions is the atomic element while in the latter scenario, single request is the atomic element and different requests issued by the same user can be directed to different web
servers. To the best of our knowledge, no research has been devoted in the user
distribution to application servers in n-tier architecture.
The paper proposes two methods to distribute users evoking similar transactions to the same servers. One is threshold of application reusibility and the other is limited buÞer sizes in each servers. Based on user profiles, the algorithms return suggestions of user distributions, the number of servers needed,
and the similarity of user requests in each server. The paper also discusses how
to apply the knowledge of existing user patterns to distribute new users, who do not have enough entries in the proßle and have no distribution suggestion,in the run time. The algorithms are also applied on a set of real data which are
derived from the access log of an enterprise ERP system to evaluate the quality of the suggested distributions. | en_US |