سیاست های مدیریت منابع کسب و کار محور برای سرورهای تجارت الکترونیک
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|8958||2000||17 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Performance Evaluation, Volume 42, Issues 2–3, 29 September 2000, Pages 223–239
Quality of service of e-commerce sites has been usually managed by the allocation of resources such as processors, disks, and network bandwidth, and by tracking conventional performance metrics such as response time, throughput, and availability. However, the metrics that are of utmost importance to the management and shareholders of a Web store are revenue and profit. Thus, resource management schemes for e-commerce servers should be geared towards optimizing business metrics as opposed to conventional performance metrics. This paper uses a state transition graph called customer behavior model graph (CBMG) to describe a customer session. It then presents a family of priority based resource management policies for e-commerce servers. Priorities change dynamically as a function of the state a customer is in and as a function of the amount of money the customer has accumulated in his/her shopping cart. A detailed simulation model was developed to assess the gain of these dynamic policies with respect to policies that are oblivious to economic considerations. Simulation results show that the multilevel dynamic priority scheme suggested here can significantly improve the values of business-oriented metrics, such as revenue per second, during peak periods. E-commerce sites that use this approach will be able to improve revenue at peak times with the same server capacity.
It has been recognized by many that congestion and poor performance can be the major impediments for the success of e-commerce. Many e-commerce sites, especially those in the financial trading business, have been facing serious problems and financial losses when customers are not allowed to trade in a timely manner. Some disgruntled customers sue on-line trading services if they feel that they have been short changed and others just move their business elsewhere. IT managers of Web stores have been managing allocation of resources such as processors, disks, and networks, by tracking conventional performance metrics such as response time, throughput, and availability. However, the metrics that are of utmost importance to the management of a Web store are revenue and profit. Thus, resource management schemes for e-commerce servers should be aimed at optimizing business-oriented metrics such as revenue per second instead of focusing on conventional performance metrics. Resource management policies for e-commerce sites should be based on the behavior of customers and on how they change state as they navigate through the site, going from browsing to searching, selecting items, adding them to their shopping carts and paying. We present in this paper a family of priority based resource management policies for e-commerce servers. Priorities change dynamically as a function of the state a customer is in, as a function of the user profile, and as a function of the amount of money the customer has accumulated in his/her shopping cart. The policies can also be tuned to provide good performance to customers who are just entering the Web store even before they add any items to their shopping carts. We believe that resource management policies such as the ones presented in this paper should be integrated into future commercial e-commerce products. This would allow e-commerce servers to handle peak loads with existing resources in a way that minimizes revenue loss due to poor quality of service. A detailed simulation model was developed to assess the gain of our policies with respect to policies that are oblivious to monetary considerations. The simulation uses the same probability distributions employed by SURGE , a workload generator for Web sites, augmented by a generator of e-commerce requests that mimic typical customer behavior. Requests generated by a customer within a session are generated from a customer behavior model graph (CBMG) that captures how users navigate through the site. The CBMG representation is used in this paper as a means of characterizing workloads for e-commerce sites. As an example, two types of customer profiles, heavy and occasional, were considered. The results of our simulations show that the dynamic priority scheme introduced in this paper can increase business-oriented metrics such as revenue per second by almost 30% over the no priority case. The rest of this paper is organized as follows. Section 2 discusses new metrics for e-commerce sites. Section 3 describes e-commerce workloads as composed of session requests and CBMGs. Section 4 describes new resource management policies for e-commerce servers. Section 5 describes the simulator and the simulation environment used to analyze the new policies proposed. Section 6 describes the numerical results obtained. Section 7 compares our work with that of others. Finally, Section 8 presents some concluding remarks.
نتیجه گیری انگلیسی
In this paper, we have introduced novel concepts for managing and evaluating e-commerce sites. We defined a set of new performance metrics to assess the performance of e-business sites in terms of business goals. Traditional quality of service models for the WWW are associated with two viewpoints: client-side performance, usually measured by response time, and server-side performance, usually represented by throughput. The metrics proposed here combine the two views. For example, revenue throughput, measured in dollars per second, implicitly represents customer and site behavior. If a customer is happy with the site service, he/she will shop at the Web store and the revenue throughput will increase. In order to maximize the revenue generated by a site and support its business goals, we proposed a dynamic multilevel resource management policy. Priorities change dynamically as a function of the state a customer is in and as a function of the amount of money the customer has accumulated in his/her shopping cart. A detailed simulation model was developed to assess the gain of adaptive policies with respect to policies that are oblivious to economic considerations. Simulation results show that for an overloaded site, an increase of almost 30% in revenue could be obtained by the priority schemes of the adaptive policy, when compared to the no priority policy. We also show that the number of angry customers that leave a site due to poor performance could be decreased by 16% by the use of priority policies. The potential lost revenue could be reduced in 50% by the priority schemes. Overall, we observed a gain in the business metrics of an e-commerce site when adaptive policies for managing resources are used. The importance of the results discussed in this paper lies in the fact that e-commerce sites that use our dynamic priority scheme will be able to improve revenue at peak times with existing server capacity. The priority techniques presented here can be adapted to handle cases in which customers can be identified by the site either through a login process or through cookies. In these situations, the site may use previously stored buying patterns to determine customer priorities. The use of a combination of current buying behavior with past buying patterns would avoid frustrating customers who have been very good buyers in the past but are mainly window shopping in the current session. It should also be noted that the policies presented here are designed to handle short term increases in traffic. In these cases, sites cannot add more resources but have to decide how best to serve a sudden load increase. One option is to refuse connections to guarantee a good quality of service to the accepted connections . One problem with this approach is that we do not know ahead of time how valuable the rejected customers might be. We opted for policies that allow everybody into the site but dynamically allocate the site’s resources in a manner that increases the overall revenue. The priority scheme suggested here requires modification to Web server, operating system, and DBMS software. While this could be seen as a drawback, many e-commerce sites already rely on open source software such as Apache and Linux. This makes it easier to implement the techniques discussed here. Developers of software for e-commerce may be interested in implementing the techniques presented here. The purpose of this paper was to present the resource allocation mechanisms for e-commerce servers. In particular, we focused on priority based processor and disk scheduling techniques and showed their effectiveness. A natural next step is to implement them in a prototype of an e-commerce site. This would allow us, among other things, to evaluate the overhead of implementing these priority schemes.