E-Book Content
GEC0N2OO6 o S Proceedings of the 3rd International Workshop on Grid Economics and Business Models • m Searching for a Provider will perform negotiation with the providers on the price. The billing components of our framework is implemented as a banking 85 agent that facilitates inter-bank communication, which consists of exchanging cheques and drafts as the means of transfer payments. As part of the testbed, we implemented a GridBank service that provides account management and credit transfer mechanisms. 3.2. GridBank Service The GridBank service is implemented as a OGSI Grid service using Globus Toolkit 3, which fulfills on the role of the banking component in our framework. Here we list the requirements that our GridBank service needs to satisfy: • Must provide account management, such as opening, checking, and closing the account. • Must facilitate money transfer between Grid Banks. • Must facilitate the query of a transactions history. • Must enable currency exchange. • Must allow the creation of account hold. The last point is enforced to assure the provider of getting paid after the service duration is over. By holding to the buyer's account, GridBank will automatically refuse any withdrawal that could reduce the balance below the amount hold. The account hold itself is not persistent and has an expiry time. Figure 4 depicts the GridBank class diagram. The GridBankingService is a persistent Grid service that satisfies all of our requirements. Furthermore, when createHold() is invoked, it will create an instance of GridBankingHoldService, which is defined as a transient Grid service. It has a well-defined termination time which can be prolonged or shortened when the need arises. 4. Related Work There are some increasing interests in applying economic approaches to Grid computing technology, but the efforts are mainly on developing economy based application schedulers 2 ' 3 and very few of them focus on providing necessary building blocks for commercialization of Grids. Computational economy enables the regulation of supply and demand of resources, offers incentive for resource owners for leasing, and promotes QoS based resource management 4 . The most notable effort of applying computational economy in Grid computing is the Gridbus project lead by Dr Rajkumar Buyya 5 . The GridBus aims at applying some economic rules for better 86 GridSeru'ce X GridBankingService ^•'Exchange Rate •,$openAccount() . %loseAccount() •\^checkAccount() • ^transferinO • $transferOut() . ^clearingln() a 2 which is type #'s participation constraint. Moreover, If (pi,P2) are chosen such that be>J > a2, then we can get the needed regularity condition that higher types get greater surplus. A simple sufficiency condition for this is that jkpl is weakly increasing in 9 (that is, y&pr > ^fy)- Under this condition, there is an indifferent type 9 such that types in [9,1] purchase the service while those in [0,9} do not. Specifically, n • t. , .. f V(0) pi(p2-0.5pi) 9 is the solution to , = — -. b{9) p2 Last, note that this co-designed contract is equivalent to the three-part tariff {pia1b(9),a1b{9),p2). 3.2. The Seller Recall that the seller can make capacity commitments at a unit cost ci and make on-demand capacity enhancements at a higher unit cost c 2 . The seller's cost function for serving customer 9 (assuming he signed the contract), conditional on capacity p, and marginal costs c\, c-i for levels below p and above p respectively, is C(p;9) = clp+^-)J (q-p)dq (7) Given the type of the consumer (given 9), the seller minimizes the above function, obtaining P* = 0ib(6) C*(6) = (32b(9) (8) (9) (10) where A = 2-^C2 C2 (ID 99 The seller's revenue function from serving