## An Equilibratory Market-Based Approach for Distributed Resource Allocation and Its . . . (1996)

Venue: | In |

Citations: | 20 - 2 self |

### BibTeX

@INPROCEEDINGS{Kuwabara96anequilibratory,

author = {K. Kuwabara},

title = {An Equilibratory Market-Based Approach for Distributed Resource Allocation and Its . . .},

booktitle = {In},

year = {1996},

pages = {53--73},

publisher = {World Scientific}

}

### Years of Citing Articles

### OpenURL

### Abstract

Resource allocation in a multi-agent system is achieved through associating agents with resources and activities. Agents associated with activities (activity agents) request agents at resources (resource agents) for the resource they require, and resource agents decide whether requests are granted. Agents are given utility functions, and each agent takes actions independently of each other to maximize its utility. Agents may not know the global state of the system, and their actions are determined based on limited information of the system. A key research question in multi-agent resource allocation is whether a multi-agent system as a whole achieves its global objective, when agents make independent decisions to maximize their own utilities based on limited information of the system. In order to investigate this question, we introduce a market model to multi-agent resource allocation. In our market model, resources are allocated to activities through buying and selling of resou...

### Citations

287 | A market-oriented programming environment and its application to distributed multicommodity flow problems
- Wellman
- 1993
(Show Context)
Citation Context ...l of (〈t − D〉,t) is same as that of (〈t − 2D〉, 〈t − D〉). In other words, the estimated price, ˆpi(t), is given by the following equation: t −〈t− D〉 ˆpi(t) =pi(〈t − D〉=-=) +{pi(〈t − D〉) − pi(〈t − 2D〉)} . (14) 〈t − D〉��-=-�〈t− 2D〉 When a new call is generated, channel bandwidth is allocated to this call. Let ∆R + j (t) denote the amount of the bandwidth required for a new call arriving from traffic source j at ... |

118 |
Enterprise: A Market-like Task Scheduler for Distributed Computing Environments
- Malone, Fikes, et al.
- 1988
(Show Context)
Citation Context ...st expensive physical connection in Sj, say, physical connection k, we use the following equation to calculate the bandwidth to request at time t, xk,j(t), xk,j(t) =xk,j(t − 1) + α(Rj − xk,j(t ��=-=� 1)). (10) Here, α-=- (0 ≤ α ≤ 1) is called a sensitivity factor. If α = 0, the above equation reduces to xk,j(t) =xk,j(t − 1), and the requested amount in time slot t becomes same as that 6sin the previous time s... |

107 |
Microeconomic algorithms for load balancing in distributed computer systems
- Ferguson, Yemini, et al.
- 1988
(Show Context)
Citation Context ... of the resource at location i. We assume that Ni is constant. Since the total demand for the resource at location i is smaller than Ni, the following equation holds: n� xi,j(t) ≤ Ni j=1 (1 ≤ i =-=≤ m). (2)-=- We assume that the value of Ni is large so that the above constraint is always satisfied. In other words, resource requests are always granted, and the requested amounts of the resource are always al... |

103 | Economic Models for Allocating Resources in Computer Systems
- Ferguson, Nikolaou, et al.
- 1996
(Show Context)
Citation Context ...e at location i, xi,j(t), also represents the amount actually allocated. Let ui(t) denote the utilization of the resource at location i at time t. ui(t) is defined as follows: �nj=1 xi,j(t) ui(t) = =-=. (3) We -=-introduce a global objective of equally utilizing resources at different locations † . In the simulations to be presented later in this chapter, we use the variance of the † In a more general case... |

100 | The behavior of computational ecologies
- Huberman, Hogg
- 1988
(Show Context)
Citation Context ... to denote a set of lj available physical connections (|Sj| = lj). Activity j issues resource requests only to these physical connections. Namely, � D xi,j(t) ≥ 0 i ∈Sj xi,j(t) =0 i �∈ Sj. 5=-= F link E (7)-=-sThe global objective is to equally utilize different physical connections. In the following simulations, time is slotted. Communication between buyers and sellers (to exchange resource prices and req... |

96 | Market-oriented Programming: Some Early Lessons
- Wellman
- 1995
(Show Context)
Citation Context ...j at time t. Assume that we allocate the resource at location i to this call. Then, the demand for the resource at location i at time t, xi,j(t), becomes xi,j(t) =xi,j(t − )+∆R + j (t) − ∆x ��=-=� i,j(t). (15) Not-=-e that t − represents the time immediately before t. Since the buyer’s strategy is to minimize its expense, buyer j estimates the current resource prices and acquires the required resource for a n... |

86 |
Markets and Computation: Agoric Open Systems
- Miller, Drexler
- 1988
(Show Context)
Citation Context ...tion k. For the other physical connections, say, physical connection i ∈Sj (i �= k), the resource amount to request is determined using the following equation. xi,j(t) =(1− α) × xi,j(t − 1) =-=(i �= k). (11) N-=-ote that the sumof the requested amounts ( � m i=1 xi,j(t)) becomes equal to the total amount of resources an activity requires (Rj). 3.2. Simulation Parameters Parameter values used in the followin... |

77 | Controlling chaos in distributed systems
- Hogg, Huberman
- 1991
(Show Context)
Citation Context ...ding at time t of a buyer associated with activity j is given by �m i=1 pi(t) xi,j(t). We define the utility U b j (t) of buyer j as a negative of the spending, namely, U b m� j (t) =− pi(t) xi,=-=j(t). (6) i=1 N-=-ote that minimizing a buyer’s spending is equivalent to maximizing the above utility of a buyer. A seller’s utility is affected by buyers’ decisions, i.e., how much resource buyers request fromw... |

30 |
A Market Approach to Operating System Memory Allocation,” Working
- Cheriton, Harty
- 1993
(Show Context)
Citation Context ... seller is to maximize its earnings. We define the utility at time t of a seller associated with the resource at location i, U s i (t), as follows: 2.2.2. Buyer’s Utility U s n� i (t) =pi(t) xi,j(=-=t). (5) j-=-=1 The objective of a buyer is to minimize its spending. Spending at time t of a buyer associated with activity j is given by �m i=1 pi(t) xi,j(t). We define the utility U b j (t) of buyer j as a ne... |

25 |
Survey of Traffic Control Schemes and
- Bae, Suda
- 1991
(Show Context)
Citation Context ...t) represents activity j’s demand for the resource at location i at time t. Since activity j’s total demand for the resource is Rj(t), the following equation holds: m� xi,j(t) =Rj(t) i=1 (1 ≤ =-=j ≤ n). (1) L-=-et Ni denote the total amount of the resource at location i. We assume that Ni is constant. Since the total demand for the resource at location i is smaller than Ni, the following equation holds: n�... |

13 |
Coadaptive behavior in a simple distributed job scheduling
- Glockner, Pasquale
- 1993
(Show Context)
Citation Context ...ify simulations. 3 Nisresource utilization, Vu(t), as a measure of how close resource utilizations are at different locations. Vu(t) is given as follows: �mi=1 (ui(t) − ( Vu(t) = �m i=1 ui(t))/m=-=) 2 . (4)-=- m The global objective is, then, to minimize the value of Vu(t). 2.2. Market Model In our market model, the resource at each location has its price. Let pi(t) denote the price of a unit amount of the... |

7 |
Symbiotic Approach to Distributed Resource Allocation: Toward Coordinated Balancing
- Kuwabara, Ishida
- 1992
(Show Context)
Citation Context ...e becomes the same as the resource amount actually allocated to activities. Therefore, using the demand � n j=1 xi,j(t), the resource utilization at location i is given by ui(t) = � nj=1 xi,j(t) N=-=i . (8) In -=-the following simulations, we use the resource utilization ui(t) as the price of the resource at location i for simplicity. Namely, we have 3.1.2. Buyer’s Strategy pi(t) =ui(t) = � nj=1 xi,j(t) Ni... |