[attachment=15236]
Abstract
Agent mediated bandwidth market, like any other market, is one where buying and sell¬ing of a commodity takes place. The item of interest is bandwidth and trading is done by adaptive, software agents, called pricebots and shopbots. In this work, a simple market model for bandwidth market is proposed, which dif¬fers from the normal market, in that it has a constant supply of its commodity. More¬over, bandwidth is non-storable and the cor¬responding market is highly sensitive to fluc¬tuations in demand and supply status. Two different market strategies are considered, one where the buyer can choose any seller for transaction and second, where the buyer only looks for the seller with the optimum price to maximize his profit. A mathematical model is constructed for market analysis and the ef¬fect of two different pricing strategies is ex-tensively studied via simulations in a variety of market scenarios.
1 Introduction
Agent mediated e-commerce is an active area of re¬search these days, especially with the exponential growth of the Internet. Agents differ from the tra¬ditional software, in that they are semi autonomous and personalized. They are usually entrusted with a fixed goal to satisfy in a fixed set of environ¬ment. Agents have the ability to communicate with other agents(peers) present in the same environment. These agents can be used in expert brokering task such as network selection, connection negotiation and bandwidth trading.
The information economies group at IBM([1, 2]) have studied the dynamics of shopbots, agents em¬ployed by the buyers to purchase goods on the inter¬net, and pricebots, agents employed by the sellers to do pricing of a commodity for sale. Dynamic pric¬ing using pricebots has also been studied by Joan Morris([6, 7]) at MIT Media lab using the learning curve approach to do online pricing of goods.
Shopbots are agents that automatically search the Internet for goods and/or services on behalf of con¬sumers. For example, acses.com compares the price and expected delivery times of books offered for sale online, while jango.com and webmar- ket.junglee.com can offer everything from apparel to gourmet groceries. On the other hand, pricebots are automated agents that employ price-setting al¬gorithms in an attempt to maximize profits. A prim¬itive example of a pricebot is available at books.com, an online bookseller. When a prospective buyer ex¬presses interest in a given book, books.com automat¬ically queries amazon.com, Borders.com and Barne- sand Noble.com to determine the price that is offered at these sites and then undercuts by 1% the lowest of the three quoted prices.
Agent mediated bandwidth market is one where pricebots and shopbots participate in selling and buy¬ing of bandwidth. Bandwidth as a commodity for sale differs from the normal commodity, in it's aggregate supply being fixed over a period of time, contrary to the normal commodities whose supply can be con¬trolled by the firms. Bandwidth is non-storable in the sense that nrmsed capacity from yesterday has no valne today. Inventories act to smooth variations in supply and demand. When no inventories exist, prices can jump if supply or demand change suddenly. Prices can also change suddenly when the perception or expectation of supply or demand status suddenly changes. Bandwidth is non-storable so price jumps and spikes (in both directions) are to be expected.
Deregulation of the telecommunications industry, advances in transmission and routing technologies, and the increasing demand for network capacity by a large number of service providers and end-users are the main factors changing the way bandwidth is bought and sold today. The development of an open and efficient market leading to optimal allocation of network resources, reduced search costs, price trans¬parency. and the development of instruments for risk management is an exciting prospect.
The issue of bandwidth pricing concerns itself with the parameter nsed to measure the consumption by the end nser. and hence to price bandwidth. Among the various parameters seen in real life situations are bandwidth page link and the connection time. In this pa¬per. a homogeneous market with respect to the type of bandwidth connection has been studied and we have concentrated on pricing based on connection time, by taking as inputs the start and end time for which the request for bandwidth is made.
Olov Schelen in her PhD thesis[5] has explored the possibilities of resource reservation on the internet to ensure Quality of Sertrice(QoS). In snch a scenario agents can effectively be utilized to price the services offered in real time, and to maximize the profit in a competitive environment.
This paper is organized as follows. Section 2 con¬structs the economic model for the bandwidth mar¬ket and explains why it is different from previously studied models. The two pricing strategies are dis¬cussed in Section 3 while simulation results obtained by varying the market parameters are presented in Section 4. Finally. Section 5 is the concluding sec¬tion. in which we disenss the feasibility of the model and approach taken in the paper to analyzing real life
Link Capacity
Bandwidth
Time T
Figure 1: Bandwidth Reservation
bandwidth markets and highlight the possible future work in the area.
2 Market Model
The Market model we have considered is derived from the market model proposed by A Grewald and Jef¬frey O. Ivephart in ([3]). This model has B buyers and S sellers. In this paper, we are looking from the perspective of the sellers and the objective is to maxi¬mize their profit. Typically. 5' -C B. Each seller has a fixed amount of bandwidth to sell. This is the seller's page link capacity L (refer to Fig 1). The implication of fixed page link capacity for each seller is that we are con¬sidering a market which is homogeneous with respect to the bandwidth connection the sellers have to of¬fer. A seller agent receives a request from the buyer agent for reservation of bandwidth b(< L) starting from time t„ to time tc. Fig 1 is a snapshot of the admitted flows in the time/bandwidth diagram. In this figure .4 is the new request that has arrived, while B .. .F has been allocated earlier. The new request is granted if the sum the new flow and the aggregate of flows from t„ and tc do not exceed the total page link capacity L. Moreover, the profit of the seller accrues both to the fraction of bandwidth requested by the consumer together with the connection time. Hence, pricing of bandwidth based solely upon the type of connection is not taken into account.