Equilibratory in MarketBased Approach Introduction Part

1
Equilibratory in Market-Based Approach-Introducti
on Part-

• Paskorn
• DSSG _at_ UMB
• 6/17/05

2
Agent in Market Model

• Activity Agent
• Seller
• Buyer

Seller
Buyer
Resource
3
Bj
Si
Si
Bj
Sm
Bn

• Seller Strategy
• Maximize earning
• Strategy
• Change price
• Bidding strategy
• Equilibratory

• Buyer Strategy
• Minimize Spending
• Strategy
• Change amount of resource based on the previous
  price

4
(No Transcript)
5
Ni total amount of resource at location i
(constant)
Assume Ni is very large. This constraint is
always satisfied
6
Utilization of resource at location i time t
7
Variance of resource utilization
m number of location (i max)
Objective of resource
8
Market Model

• Pi(t) price of unit amount of
• resource at location i time t

is changed based on its past demand
9
Sellers Utility

• Max earning

10
Buyer Utility

• Buyer determines amount of resource based on past
  price.
• Objective minimize spending

11
Case Study
n number source destination pair
(activity) i bandwidth of physical
connection Ni total amount of bandwidth for
physical connection I Rj total amount that
source-destination pair j requires Sj set of
paths that source-destination pair j exists
12
Agents Strategy

• Sellers Strategy
• Change price
• Buyers strategy
• Change amount of required resource

13
Sellers Strategy

• Since, every required resource always meet

Thus, we can set the price from the resource
utility
14
Buyers Strategy

• Determine Xi,j (t) (demand)
• Required amount of physical connection is
  bandwidth
• Let k resource location i that has the least
  expensive price

15
For the least expensive physical connection

• alpha (0 1)
• 0 then
• 1 then

The amount of bandwidth at the least expensive
physical connection is the same as its past
The required resource is requested from the least
expensive physical connection (k)
16
For the other physical connection
17
Simulation

• m 100
• n 100
• Rj 50,000 (total amount of resource that
  source-destination pair require) (constant)
• Ni amount of resource at location i (the
  bandwidth of physical connection i)
• is uniformly distributed between 1 to 1,000,000
• average of Ni 500,000
• The average of resource utilization 0.1
• Assume 5 physical connections for each
  source-destination

18
Result
slow
Fast but swing
Equilibrium value
19
Result
fast
slow
20
Conclusion

• Oscillation was cause by a large sensitivity
• Small values of sensitivity reduce the degree of
  oscillation
• Agent can be added learning algorithm to change
  the price or required resource