JP2002540510A - System and method for implementing a progressive second price auction approach - Google Patents

Abstract

(57) SUMMARY A system and method for allocating resources uses a progressive second price auction approach. The auction is held on a limited resource, such as the bandwidth of an Internet service provider network, and the bid is made by a prospective user (103) having a desired amount and a price for each unit of the bidded resource. Done. To make a valid allocation, the new bidder (105) allocates some of the resources based on the availability of resources (109) constrained for higher bids than the new bidder (113). To give permission. The actual price paid (121) by the new bidder is based on the lower priced bid to which some of the resources have been or will be allocated. With such a calculation of the price paid, the bidder bids on a valid valuation of the resource relative to an inappropriate bid contract.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[Field of Technology]

The present invention relates to a system and method for performing a progressive second price auction to effectively allocate finite resources among competing requesters.

[0002]

[Background Art]

Significant problems arise when a finite separable resource is desired by more than one person and more resources are required than can be allocated and distributed to the requesting entity. If demand for resources or products exceeds supply,
A market auction may be held to properly allocate the restricted resource or product to the requesting entity. For example, finite commodities such as oil are produced in a limited amount per month, and oil is auctioned and sold in international markets. If only one million barrels of oil are produced per month, bidders can bid through international markets, typically with the highest bidder winning the requested oil quota and providing oil. Pay the bid price.

[0003] The problem of allocating finite resources also occurs in the electronic world. The Internet allows bidders to communicate and exchange data over long distances. However, the rate at which information can flow can be affected by the number of users using an Internet Service Provider ("ISP"), forcing users to avoid prolonged pauses when connecting to a network. . If important data needs to be transmitted and received almost instantaneously over the Internet, the ISP's premium bandwidth can be reserved for that user. If premium bandwidth can be provided by the ISP, then bandwidth becomes a constrained resource and users want all or part of the available premium bandwidth. A standard auction of bandwidth consists of allocating premium bandwidth to the bidder who pays the highest bid price. However, this type of auction is inefficient for a large number of bidders, in which case the bidding and availability of premium bandwidth varies over time. Other allocation techniques require exchanging large numbers of messages to carry the required information and are not suitable for communication networks due to heavy signaling loads. An improved auction and allocation approach is needed.

[0004] To achieve an improved assignment approach, some economic principles can be applied to the approach. An example of the allocation of premium bandwidth resources is used in the description below. The value of communication networks like the Internet, the external effects of economists
(externality). According to the principle of external effects, the value that a user obtains from the network depends on the evaluation and behavior of other users. According to certain recognized principles, when more people are connected to the network, the communication network becomes more valuable to the user. According to a second recognized principle, as the usefulness of the network for one user increases, the quality of service obtained by other users decreases. Resources are allocated by users who cannot adjust their operations sufficiently to make the most desired allocation of resources due to distance, group size or individual egoism. This external effects principle model shows that the auctions of other bidders can improve the allocation of resources on the communication network using a game-theoretic approach to determining allocations and costs to bidders.

[0005] The allocation of the limited resources associated with the communication network is used to obtain an example of a real market where a user bids on actual dollars or other items of value and to obtain the limited resources of the processing entity. And a private system example of bidding using internal bucket assignments. For example, different departments of a university bid using the university money provided by the university on the time used by the components of a particular university.

[0006] In a new multi-service network (ATM, next-generation Internet), uniform pricing based on the peak capacity of user connections as used in the current Internet, and time-based prices (time-based) as used in telephone systems. of-day usa
ge price) is not a valid solution to make a valid allocation of resources. Therefore, it is necessary to develop a new method for setting network resource prices. Effective pricing system requirements are: (1) resource requirements (by a sufficiently wide range of traffic classifications priced differently or by allowing users to clearly quantify resource requirements); Range sensitivity, (2) the need for prices to dynamically respond to unpredictable demands (system-based markets), and (3) the possible trading constraints for which a pricing architecture is possible. Including needing to be smaller. It is preferable to implement an auction approach that meets the above requirements and provides a more efficient allocation of finite resources.

[0007]

DISCLOSURE OF THE INVENTION

It is an object of the present invention to provide a system and method for allocating separable resources using a progressive second price auction technique. The method retrieves data representing at least one previous bid having a quantity data component and a price data component, retrieves data representing a new bid having a quantity component and a price component, and retrieves a higher price than the new bid. Allocate resources to the first bid with a component, allocate resources to a new bid if some of the resources are not yet allocated, and to the price components of other bits that are not allocated if there are no new bids. Calculate the cost of a new bid based on it.

[0008] By calculating the cost of a resource based on the principles of a progressive second-price auction approach that calculates the cost to the bidder in response to other bidders, each bidder can bid with a fair assessment of the resource. And avoid using inefficient bidding procedures. By using this method, a more effective allocation system is obtained. Process other new bidders and base the allocated resources and allocation costs on the bidders on other bids.

[0009] One example of a resource that can be allocated is premium bandwidth by an Internet Service Provider (ISP). The ISP
A second price auction approach is used to process bids from different clients requesting premium bandwidth. In one example, the ISP makes an initial assignment based on the first bid and processes each new bid before the first assignment is made. In the second example, no such initial assignment is required. Bids during an auction are categorized by price component, allocating resources to bids with a price component above the new bid and allocating resources to the new bid. The cost of a new bid is based on the bidder who placed the lower bid, allocating some bandwidth if no new bid was made. Thus, the cost to the bidder is based on other bids. The cost charged for a new bid is
It is equal to the sum of the reduced quotas for other bidders weighted by the bid price of such bidder. This approach of calculated payment, or other bidding, encourages people to bid on the actual value for the resource. The reason is that the cost paid by the winning bidder is based on the bids of other bidders. The ISP may also set its own minimum price for a bid given an assignment to pay a predetermined cost level.

[0010]

BEST MODE FOR CARRYING OUT THE INVENTION

The method of the present invention is a progressive second price auction.
second price auction). The second price auction is performed as follows. When an object is auctioned between participants, the highest bidder is given the object, but the winning bidder has the second highest price. This type of bidding makes it easier for bidders to bid on the item with an accurate valuation, compared to trying to bargain for inefficient bids, such as $ 1 more than the highest expected bid. Is performed efficiently. Since the highest bidder pays the cost of the second highest bid, it will not be bid lower than the actual rating of the bid. If all other bidders are well below the bidder's bid price, the bidder need only pay the next higher bid price. If a bidder bids above the item's rating, the bidder pays a higher bid value than that rating, depending on other bidders he does not want. Therefore, they will not pay a higher bid value than the actual evaluation.

[0011] The applications of the progressive second-price auction approach range from the example of a single entity to bidding on collateral for separable resources that have expired. Such an approach is a very effective approach for auctioning and allocating finite resources among multiple bidders. With multiple bidders in the auction, most participants receive a portion of the resource that is equal to the rating of the resource being auctioned.

FIG. 1 is a flowchart of the steps for performing a progressive second price auction technique. A general resource unit auction will be described while auctioning any resource or other separable goods or services using the technique shown in FIG. An application for an actual resource that can be allocated is, for example, a premium bandwidth. During FIG. 1 and throughout this application, a bid is defined as having two components: a quantity component and a price component. For example, a bidder during a resource auction bids 5 units (amount) at a price of $ 2 per unit. Other components or other information may also be included in the bid as needed, only while the amount and price components of the bid are described in the preferred embodiment.

Step 103 retrieves a list of previously placed bids for a particular auction resource that previously classified the decreasing bid price. This means, for example, that the highest bid offered will be first in the list. Bids can also be categorized after a search. The list of bids is stored in a computer file in the memory of the processor performing the auction, entered by a keyboard, transmitted in real time over the network with the bidder's software bid approval during the auction, or performing a progressive second price auction procedure. The search can be performed by any other means. For example, if ten bidders have previous bids for the auction's premium bandwidth, these bids (including the desired amount and the price per unit provided) will be available on the computer running the method or It is stored in the memory of another device. The bids are then sorted by conventional sorting techniques, placing previous bids in order from highest to lowest. Bids may be stored in a linked list upon receipt, in order of decreasing price. If the bit to be processed is the first bid made during the auction and no other bits are stored, the bid will be the first bid in the list.

Step 105 searches for a new bid (S _i ) to process having at least a quantity component and a price component. The quantity component is in a predefined unit for the auction, for example, in 1 Mbps (megabits / second) of bandwidth, and the price preferably takes the form of a price per unit. However, the price can also be entered in the form of a full price for a quantity bid, and the unit price can be easily calculated. Using a second-price auction approach, the new bidder optimally places a bid that is the bidder's near-valid assessment of the resource. Since the second price auction technique is performed as described above, the bidder optimally bids at that legitimate value. New bids can be made electronically over a network such as the Internet, or can be manually entered into a computer that performs the procedure. For example, a new bidder who wants premium bandwidth may have a bandwidth of 3
Type into the computer with a bid of $ 4 per unit per unit and transmit the bid to the computer performing the procedure over the network. Software agents can be programmed to bid automatically according to user needs and markets as represented by other user bids.

Step 107 is the total amount of resources available at a time (Q
Search). For example, if the resource to be auctioned is premium bandwidth, the total premium bandwidth for a given ISP system or node may be 20 Mb.
ps. Since there is only a finite amount of premium bandwidth resources that can be used at one time, the amount of bandwidth over Q cannot be allocated. The quantity Q can be manually entered or retrieved from electronic storage;
Alternatively, it can be calculated by an auction processor. Further, Q may be a predetermined value in a manner for the resource, or may be calculated based on other parameters. For example, if the resource allocated is bandwidth, the ISP may monitor the price of premium bandwidth available for auctioning over the entire bandwidth for a given period. The ISP can increase the percentage of bandwidth that is the designed premium. The value Q varies over time due to technical or market variables.

Step 109 first sets the remaining quantity variables (Q _i ) to be equal to the overall quantity Q. At this point in the approach to handling new bids, bandwidth is
New bidders will not be assigned by this method. The actual bandwidth,
It can be used according to the previous assignment when processing a new bid. The remaining quantity variables Q _i represent the amount of resources that were not allocated during the allocation procedure for a new bid.

Step 111 identifies the next highest bid (S _j = (Q _j , P _j )). In this case, let S _j be the j-th bid, Q _j be the j-th amount bid, and P _j be the j-th price bid. Initially, the auction technique performs this step, with the second highest bid identified being the bid with the highest price included in the bid list. Since the bid is stored by lowering the bid price, the first bid in the list will initially be the highest. If the list is implemented as a linked list, the pointer points to the highest bid.

[0018] step 113, a bid that has been currently identified in the list to check whether or not a new bid S _i or more in the auction. If the currently identified bid in the list is greater than or equal to the new bid, the method continues at step 115. If the currently indicated bid in the list is less than the new bid price, the method continues at step 117. Step 115, the currently identified Quantity Ingredient bid in the list is subtracted from the value of the available amount Q _i. Since the price of the currently identified bid in the list is higher than the new bid, the amount of the requested resource is given (to the extent that the remaining amount of resources has not yet been allocated during this allocation for the new bid). Moreover high amount of bids to the allocated resources in the bid list is subtracted from the resource variable Q _i available. If the new bid is below the previous bid from the bid list requesting an overall quota above the service provider's available quota, the new bid is not assigned to any of the resources. The reason is that resources have been allocated to higher bidders. The new bidder is forced to increase the bid for the next bid or obtain resources from other sources. Bids that are currently identified in the retrieved list of bid, to require more than the current available resource as identified by Q _i, the entity performing the bid is allocated to the rest of the resources. In this case, the new bidder receives a portion of the requested resource.

Step 116 checks whether the remaining quantity variables Q _i are equal to zero. If it is equal to zero, not assigned more resources for a new bidder S _i, to complete the method for the new bidder S _i about bids made. The bidder increases the bid price and makes a new bid. If Q _i is not equal to zero, continue checking the bids received in the list and the method proceeds to step 111. If the bid price of the new bid S _i such that with a portion of the amount assigned to the bid price is high enough, to reach the step 117. For example, with an overall quota of 9 units, the auction computer may calculate ($ 3,7), (3,
$ 6), have previously received bids (amount, price) of ($ 3,2), and
If the current bid S _i is ($ 3,5), three units are allocated to the first two bids of $ 7 and $ 6, and three units remain allocated (Q _i = 3 ). Step 117 allocates the appropriate amount of resources to the new bidder _Sj by retrieving the smaller of (1) Q _i (remaining amount) and (2) q _i (requested amount). . Therefore, the new bid S _i, is assigned the amount requested is not already assigned with available. In the above example,
Three units are assigned to the new bidder.

Steps 119-123 calculate prices for new bidders according to a progressive second price auction technique. Step 119 places the new bids in the list of bids in sorted order. For example, bids are added to a linked list so that bid prices are arranged in descending order. The lowest bid that previously received the resource assignment is identified. If the initial allocation has not yet been performed, the lowest bid to which a portion of the resource was allocated in the absence of a new bid is identified.

Step 121 calculates the total cost for the new bidder based on a progressive second price auction approach. The cost is based on the exclusion compensation principle (exc
It is calculated based on the lusion compensation principle). The new bidder i is i
Make a payment for the quota so as to cover the "social opportunity cost" given by the declared willingness to pay (during the bid) of the user who has been excluded due to the presence of the. In this way, the operator and supplier of the resource at the time of the auction guarantee the maximum lost potential revenue from the new allocation.

Thus, in the above example, if the lowest allocation to receive the resource allocation was $ 2 for three units before the new bid, these three units would be reduced to $ 5 for these three units. Is reassigned to a new bidder who bids on. The new bidder _Sj is charged $ 2 for each unit assigned. If the new bidder makes a bid of ($ 3,5.5), the stored bid list will have the first three bids each assigned 3 units ($ 3,7), (
$ 3,6), ($ 3,5) and ($ 3,2). The $ 5.5 new bidder is allocated three units of $ 5 bidder at a cost of $ 5 per unit according to the second price auction rules. Thus, the quota price is always based on the bid of the lower bidder who had the quota previously received or who received the quota if there were no new bidders.

Step 121 shows how to calculate a cost for a new bidder to receive the resource allocation. For each of the lower bids that received the resource allocation before the new bid, the bid price will be charged to the new bidder until the allocated amount matches the bid requirements. Added to cost. This reflects a shift in allocation from low bidders to new bidders who have bid higher. The amount reallocated for each of the previous bids is responsive to the price at which the amount was previously allocated, ie, has that allocation level. Thus, the cost C _i is equal to the sum of the smaller of the re-allocated quantity (1) and the quantity allocated for the previous bid (2) multiplied by the unit bid plus C _i. Become equal. If reallocating a two-unit allocation at $ 3 while processing a four-unit new bid at $ 4, two units are reallocated from the lower bid to the new bid. The amount allocated in the next highest (but lower than the new bid) bid is used to further calculate the total cost of the new bid reflecting the replaced allocation.

[0024] After performing the reallocation for a particular bid amount to be reassigned for the new bid, the minimum of the amount of allocated bid Q _i retrieved to the amount a _i to be reallocated It is decreased as the _{(a i = a i -m w} (j, a i)). Thus, once the requested quota during a bid is taken into account by the assigned or assigned one or more lower bids, no further cost considerations are required.

Step 123 is a calculation C_jNext higher P to use in_jIdentify. Then
The process continues at step 125. Step 125 is a_i(Reassigned
Whether the remaining amount) is zero or P_iGreater P_jCheck for the presence of
I do. If one of these conditions is true, processing is complete for the new bid.
Complete. Other low P to process_jExist and some of the
If resource exists, process P_jIs continued at step 121. FIGS. 2 and 3 are diagrams of changes in resource allocation in response to a new bid to be processed.
This is a typical example. Figure 2 shows a new bidder running an auction to process new bids.
4 illustrates the allocation of resources based on previous bids prior to the start. In this example, the total amount of resources
(Q_T) Is 35 units as shown in a box 203. In this example,
The bill is retrieved as represented in box 205. First Searched Bid
S₁For 7 units of $ 1 per unit, the second search bid S ₂ For 18 units of $ 2 per unit, the third search bid S ₃ For 5 units of $ 3 per unit, and the fourth searched bid S₄ To 10 units of $ 10 per unit. The graph 201 shows the bid amount 207.
4 shows a horizontal axis and a vertical axis representing a bid price 209. Bid placed to right of vertical axis
Is assigned a portion of the auctioned resource. Bidding first, the highest
Assigned according to bid price. Therefore, the fourth bid S₄(Max bid price
) Is represented by a bar 211. Bar 211 shows the 10 requested by bid.
Indicates that the entire unit has been allocated. Bar 213 displays all required 5 units
3rd bid S that received₃Shows the assignment to Bar 215 requested
Second bid S received all 18 credits₂Shows the assignment to Bid S₄,
S₃And S₂Is assigned, only two units are not assigned (
35-10-5-18 = 2). Therefore, the bar 217 displays the bid S₁Percent
2 units assigned and S not receiving quota (shown to the left of the vertical axis) ₁ Shows the 5 units required by. The price level on the graph represents the bid price
Does not represent the price actually paid. The cost of a successful bid is related to Figure 1.
Calculated using the progressive second price auction method described
.

FIG. 3 shows a graphical example of a method of changing the assignment when a new fifth bid is made on the auction as S ₅ ($ 5.25) by this method. Figure 3 shows a graph 301 having a new allocation due to the processing of a new bid S _5.
Graph 301 has a horizontal axis representing bid volume 307 and a vertical axis representing bid price 309. The total level of available volume is 35 units, as shown in box 203. The new bid S _5, together with the shown in the bid list 305, arranged in order of price bid. Bid _{S 5} with $ 2.50 bid shown between _{S 2} and _{S 3.}
In this case, the graph reflects the assignment for each bid after considering the new bid.

In FIG. 3, the entity having bid S ₄ is also allocated 10 units, as indicated by bar 311. Those with assigned S ₃ also has 5 units allocated as shown by the bar 313. Next highest bid is S _5, the bidders S _5, are assigned the requested 5 units. Between the S _{4, S 3} and _{S 5,} are assigned the 20 units, so that only 15 units remain (35-10-5-15).
Next highest bid _{S 2} requests the allocation of 18 units. However, since only 15 units do not remain allocated, _{S 2} receives a 15 unit by the bar 315. This is three units less than in the case shown in FIG. The bid number requested for seven units is assigned zero units, as indicated by the bar 317 on the left side of the horizontal axis.

The cost of the new assignment for the new bidder S ₅ will be the quantity 321 replaced by the reallocation. Therefore, the total cost for 5 units replaced with reassigned by bidding S ₅ is (3 units × 2 dollars) + (2 units × 1 dollars) = 8 dollars. The unit cost for the _fifth bidder S5 will be the total cost divided by the number of units assigned, ie, 8/5 = $ 1.6. Thus, the fifth bidder pays $ 1.6 per unit, despite bidding $ 2.5 per unit according to the progressive second price auction approach.

The auctioneer can offer its own minimum bid to purchase the resource at a particular price, as the actual price assigned and paid is based on other payments for the resource. This prevents the first bidder from acquiring resources at a very low price and allows the auctioneer to influence the price paid by others. The minimum price needs to be set to an appropriate level, or the auctioneer may not sell any resources to other bidders.

A progressive second-price auction by allowing bidders to modify bids and allowing bidders to specify a requested amount with resource prices subject to market conditions and efficiency constraints The technology meets the requirements of an effective pricing system.

To bid effectively in an auction that runs continuously and quickly, such as in auction bandwidth, an evaluation method is used to quickly bid the bidder in an electronic auction and obtain a valid result. Can be obtained. Each bidder i has a resource rating θ _i (a _i (s)) ≧ 0, which is the total value of the received allocations. A bidding profile can be formed to obtain the utility U _i (s). In this case, U _i (s) = θ _i (a _i (s)) − C _i (s), which is the value minus the cost. Further, the bidder may have budget constraints on the total cost used for the resource during the period, so that the total cost of the bid needs to be below budget parameter b.

FIG. 4 shows a utility function (utili) of a bidder when conducting an auction at a specific bid price.
4 shows a graph 401 of ty function). In the auction of this example, the following bids S ₁ = (100, ₁ ); S ₂ = (10, ₂ ); S ₃ = (20, ₄ ); S ₄ = (2
There are five other bidders with (0,7) and S ₅ = (30,12). The new bidder S ₆ has an evaluation function of θ ₆ (q) = 10q (ie, $ 10 per unit). The upper pedestal of the graph 401 indicates where the amount requested by the new bidder no longer rises from the bid price because it has been allocated to other bidders. The graph 401 has three axes: quantity 403, price bid 405, and utility 407. If the new bid is below 10, the utility increases as shown. However, the price of a new bid may be
Having more than two price components reduces the overall utility for each of the additional units. The reason is that if a bidder is willing to pay for a resource allocation, it will usually be $ 10 per unit. Another form of the utility graph is shown in FIG.

FIG. 5 shows an assignment graph similar to the graphs shown in FIGS. 2 and 3. Box 5
01 indicates the evaluation function 10q (ie, 10 times the amount), where the amount is shown on the horizontal axis and the utility is shown on the vertical axis. In the graph 503, the bid described in the example of FIG. 4 is shown. The horizontal axis shows the allocated quantity 505 and the vertical axis is the price 507 of the bid. The total amount to be allocated is 100 units. The auction bid is represented in box 509. It represents an assignment for bid _{S 1} bar 511 represents an assignment for bid _{S 2} bar 513 represents an assignment for bid _{S 3} bar 515 represents an assignment for bid _{S 4} bar 517, a bid _{S 5} bar 519 Represented by Bid _{S 5} is assigned only 20 units out of the total units. Box 50
To determine the optimal bids by the new bidder S ₆ having an evaluation function shown in 1, by utilizing the derivative of the cost function to determine the slope of the request. The derivative of the new bidder's evaluation function is shown by line 521 in graph 503. Each quantity represented by a bar below line 521 is more valuable to new bidders than to the quantity bid by other bidders having lower price components. Thus, the new bidder will have 20 units on line 513, 20 units on line 515, and line 517.
It is desired to bid $ 10 to obtain 70 of the total units, including 10 units of, and 10 units of line 519. Other units cost more than calculations on the unit, so that higher bids are not valid. The actual cost paid by the new bidder is calculated using the technique described in connection with FIG. The 30 units indicated by bar 511 are bid at a price of $ 12 above the price of $ 10, and the utility is gained by a new bidder. Thus, the new bidder gets 70 units for a bid price of $ 10. Although the derivative of a straight line is illustrated,
An accurate bidding procedure can be modeled using a more complex evaluation function.

An auction as described herein is applicable to settings where resources are arbitrarily separable. Thus, this approach is well suited to systems that can request and obtain any amount of resources, such as the pricing of the “expected capacity” profile on the Internet with different services.

For time-based network resource allocation, bidding is based, for example, on a flow or per unit time on a suitable time scale. The flow-level model can be used especially when the entity in the auction is a software bidding agent embedded in a software application, which is an accurate assessment of the utility function θ of the entity, understanding of relationships between resources or understanding of services Evolve levels (using artificial feedback over time).

One example of a flow-level auction has an agent-based system and ISP seller implemented in a common programming language, the network of which is:
It has a narrow bandwidth (maximum resource capacity) of 45 Mbps. The above example is an actual auction performed using a progressive second price auction technique. Using the flow level service model, the ISP allocates capacity to all users. In such a service model, the assignment is a traffic profile for each user made by a profile meter at the point of attachment of the network. The network provides premium services (eg, low loss and / or delay) for all in-profile packets, and any out-of-profile traffic gets the best service. Providers need to have a network that supports all the expected in-profile traffic at a given time, whereas the worst case has more capacity (
Do not sell the profile "sum". Depending on how strictly the ISP adheres to the contract and how well the ISP predicts the route to predict the traffic to carry, the ISP may determine the bottleneck capacity.
sell a total capacity of more than (city). In this example, the ISP is eventually saved, so that the ISP will generate 45 Mbps to generate all units sold.
Only sell. In addition, the ISP needs to purchase a sufficiently large profile from the neighboring network so that after leaving the network, the in-profile traffic of the user still gains connection services. The minimum auction price P ₀ corresponds to the price paid by the network that acquires capacity from the neighboring network.

The bidder's profile in this example can be any traffic descriptor. For example, quotas can provide users with a minimum guaranteed bandwidth (to the allotted amount), and each user is allowed to burst transfer to a physical line rate. This can be called, for example, a "leaky bucket" (leaky
-bucket) can be done by traffic meter. Bucket size is IS
If fixed by P, the user is assigned (and paid) a token rate. The bucket size determines the maximum in-profile burst size, which corresponds to the worst case buffer capacity or delay that the ISP can tolerate for premium traffic. Obviously, the quality of service for most Internet traffic will be better than if the allocated premium bandwidth were increased. However, since the profile allows bursts, the quality improvement decreases as the premium bandwidth approaches the physical line rate,
Of course, it cannot be used if it has a higher premium bandwidth than the peak rate.
Thus, an ideal model of a bidder's rating would strictly increase, be concave up to the peak rate, and flatten above the peak rate (bid rating curve). To represent a realistic range of connection speeds for users requesting premium bandwidth from the home or small business users to a peer network connecting using a T3 line, the user population in this example is the line rate q _i
∈ {256 Kb / s, 1.5 Mbps, 45 Mbps}. The ISP charges a flat rate T1 connection at $ 500-1500 per month. this is,
$ 0.008-0.023 / Mbps every minute. Similarly, dial-up users are charged a flat rate of $ 10-30 per month for a 30 Kbps connection, which is $ 0.007-0.021 / Mbps per month. The closeness of these two cases indicates that these flat rates are a rough measure of the average rating. Auction operators may also charge a small fee to process each bid.

The following bidding method can be used as a procedure by entry bidding in a progressive second price auction. If the utility to the bidder is increased by increasing the requested bid price or volume and remains present in the bucket, it follows the new bid with the increased price and / or volume.

The order of the bids is not important for the bidding approach. Bidders participate in auctions at different times and bids, for example, once per second. This makes the distributed auction asynchronous, according to the communication delays that make up the time for the bid to reach the server and the time to update the client at almost any time.

The bidding method can be described as both selfish and short-sighted. It is (by more than the fee for a bid)
It is selfish because its usefulness follows new bids and does not take the form of an auction, i.e. the consequences of systematically using other users rather than procedures that may temporarily decrease. Is myopic because it uses good utility.

In this example, it is assumed that the total capacity Q is 45 Mbps. The bid fee is the fee paid to the auction operator who places the bid. It is typically a lower price than the bid price. In this example, it is assumed that the bidding fee is $ 0.00125. All times in this example are expressed in seconds. In this example, the auction is conducted by one agent (bid computer) for the user, and the agents are evenly distributed to different workstations. The auctioneer agent (processing computer) runs only on a separate workstation, such as a Sun Ultra-2 SPARC workstation. The "appearance" of a new auctioneer corresponds to an Internet user whose profile is zero (pure maximum performance) and who has not placed a bid that suddenly requests a bid profile. This may be a user who is satisfied with maximum performance traffic, for example, at night and on the weekend, but demands better bandwidth during work.

In this example, (1) 20 users (bidders 1-20) having a peak rate of 1.5 Mbps
(2) Two users with a peak rate of 45 Mbps (bidders 21-22)
(3) 50 users with a peak rate of 256 Mbps (bidder 23-7)
2) Start at t ₀ <0 with the group consisting of The evaluation of each bidder follows a bidder evaluation curve as described above. Users of this set will follow that rating until equilibrium is reached at a given time t <0 (ie, no new bids will be received at a new time). The effect of a new participant who bids to enter and exit the auction is shown in FIG.

FIG. 6 shows three graphs showing the effect of a new small user on the auction. Graph 601 shows the required amount of bandwidth allocation for a 256 Kb / s line rate user. The horizontal axis of graph 601 represents time, and the vertical axis represents bandwidth requirements. Graph 603 shows the assignment for the new bidder. The horizontal axis of the graph 603 represents time, and the vertical axis represents the allocated bandwidth. Graph 605 represents bidders versus bandwidth and price paid. The horizontal axis of the graph 605 represents time, and the vertical axis represents dollars (per minute × Mbps). The bid price is represented by a solid line, and the price paid is represented by a dotted line. Since the progressive second price auction procedure is used, the line of fees paid (cost) will be lower than the bid price. Since a relatively large number of bidders are participating in the auction, the bid price and the price paid are close to each other.

FIG. 7 shows three graphs depicting the impact of a new small user on the auction from different perspectives. Graph 701 shows the required amount of bandwidth allocation for a 1.5 Mbps line rate. The horizontal axis of graph 701 represents time, and the vertical axis represents bandwidth requirements. Grab 703 indicates the actual assignment to the new bidder. The horizontal axis of the graph 703 represents time, and the vertical axis represents the allocated bandwidth. Graph 705 shows the bid for bandwidth and the price paid. The horizontal axis of graph 705 represents time, and the vertical axis represents dollars (per minute divided by Mbps). The bid price is represented by a solid line, and the price paid is represented by a dotted line. Using a progressive second price auction, the price (cost) line paid is lower than the bid price. Since a relatively large number of bidders are participating in the auction, the bid price and the price paid are close to each other.

FIG. 8 shows three graphs depicting the impact of a new small user on the auction from different perspectives. Graph 801 shows the required amount of bandwidth allocation for a 45 Mbps line rate. The horizontal axis of graph 801 represents time, and the vertical axis represents bandwidth requirements. Grab 803 indicates the actual assignment to the new bidder. The horizontal axis of graph 803 represents time, and the vertical axis represents allocated bandwidth. Graph 805 shows bids for bandwidth and prices paid.
The horizontal axis of graph 805 represents time and the vertical axis represents dollars (per minute divided by Mbps). The bid price is represented by a solid line, and the price paid is represented by a dotted line. Using a progressive second price auction, the price (cost) line paid is lower than the bid price. Since a relatively large number of bidders are participating in the auction, the bid price and the price paid are close to each other.

The following results of this example are reflected by changing the assignments for different bidders as shown in FIGS. (1) At t = 0, a participant 73 (a new 256 Kb / s user) participates. First, the new user assumes that the market is empty, so that the new user requires a large profile. Once the knowledge of the new user's opponent's profile has been updated (received by the auction processor), the new user recognizes that this is very expensive, reduces its volume, and reduces the bid price. To raise. FIG. 6 shows that after three bids by new participants (and reactions by others), the market has stabilized at a new equilibrium at t = 170, where:
Bidder 73 has an allocation of premium bandwidth a ₇₃ = 128 Kb / s. (2) At t = 550, a participant 74 (a new 1.5 Mbps user) participates. FIG. 7 shows a ₇₄ = 728K after two bids by participant 74 at t = 600.
Indicates that a new equilibrium of b / s has been reached. In this new equilibrium, participants 7
The unit price paid by 3 is 0.0 (although hardly visible in FIG. 6).
It increases slightly from 0625 to 0.00629. (3) At t = 890, a participant 75 (new 45 Mbps user) participates. Unlike the previous two cases, such a large occurrence has a significant impact on the market. FIG. 8 shows that after 28 bids of a new bidder at t = 1200, a new equilibrium of a ₇₅ = 19 Mbps has been reached. At the new equilibrium, the ownership of premium bandwidth bidder 74 drops from 728 Kb / s to 540 Kb / s,
The price paid per unit rises from 0.0063 to 0.008. (4) Between t = 1600 and 1700, 50 new users with a peak rate of 256 Kb / s join the game. This causes the market to fluctuate for several minutes, reaching a state of equilibrium where prices have risen slightly, and quotas for existing bidders have decreased slightly (see FIGS. 6-8). (5) Just before t = 2500, participants 1-20 (all at a line rate of 1.5 Mbps) suddenly leave. This causes a rapid descent in the market, which takes about 200 seconds to reach a stable state. The next equilibrium is not much different from the previous steady state in terms of both quota and price, even if the potential demand is reduced by about 30 Mbps. The reason is that there are many low-rated users who were not satisfied enough before using up to empty capacity, and the clearing prices in markets that keep prices at almost the same level
There are sufficiently unsatisfactory demands below (clearing price). (6) At t = 3140, the participants 21-22 (all 45 Mbps rate) suddenly leave. The empty capacity is sufficient to satisfy a large number of bound demands, so that the price drops to about half, about $ 0.004, and the quota increases for the remaining participants.

Over the present example, which described three participants, the T3 user (bidder 75)
Pay a lower price per unit than the other two. Such a “capacity discount” is due to the fact that there is more bandwidth allocation for the participant 75 and lower priced bidders are replaced by more required quantities in a progressive second price auction approach. Occurs.

These examples show that the impact of multiple independent small users is significantly smaller in terms of market dynamics than the impact of a single user with the same total requirement. For each small user, it is equal to the retail market because the small user cannot "run" the market. This implies that for the smallest user, the market can be approximated by a mere advertised price (due to the interaction between the medium and the large user), and the user decides whether to buy or not. Easy to choose. This occurs in most situations in practice (eg oil).

[0049] Progressive second price auctions generalize the key characteristics of a traditional single indivisible auction to the point where any separable resource should be owned. An assignment technique that assumes an adaptive requirements model of the user's preferences is called Vi
Configure a stable and effective allocation and pricing approach in the network context for both rtual Path / Virtual Network applications and edge capacity pricing applications.

FIG. 9 is a diagram of an example of a system that can use the progressive second price auction method. The system 901 includes an Internet service provider ISP1 903, ISP2 905, and ISP3 907. Connect each Internet service provider to the network. ISP1
903 to the network 909 and ISP2 905 to the network 91
1 and the ISP2 907 to the network 913. The bidding entity 915 is connected to the network (each of networks 903, 905 and 907)
, Which participates in the auction by sending an electronic bid having a quantity request component and a bid price component to an ISP connected thereto. The bids may be individual bids, or may be in the form of valuation data represented by a plurality of price component and quantity component bids, ie, a price function expression. The bidding entity 915 has a central processing unit and a memory that stores a set of instructions for formatting a bid and sending a bid to an auction processor (in this example, an ISP). The ISP updates the auction with the bidding entity 915
Can also be sent. System 901 can be used by an ISP to auction premium bandwidth.

The Internet is composed of a plurality of interconnected networks, and messages are transmitted evenly from a sending participant to a receiving participant. Thus, when a bidder places a bid on premium bandwidth, the message typically needs to travel across multiple networks to reach its destination. Therefore, premium bandwidth must be allocated by each network, and the bidding required to secure the overall connection contributes to the overall cost.

For the auction of this example, each ISP is required by a computer having a processor and memory, and will be described in more detail with reference to FIG. If the bidding entity wants to send information to an entity that is not directly connected to the sending entity's ISP, the ISP needs to reserve the appropriate bandwidth from the intervening destination ISP. This is accomplished by the ISP that is the bidding entity itself (as shown by box 904), in which case the ISP bids on other ISP bandwidths to ensure that data flow bandwidth is allocated across the connection. , Not directly through the connection of the ISP. Therefore, ISP 903
Acts as a bidding entity 904 for auctions performed by ISP2 905 and auctions performed by ISPs 907 when bandwidth from ISP 907 is required by the ISP's auction bidder. Other arrangements have the bid itself at the bidder's position in each of the ISP auctions. However, if the network path has multiple ISPs that need to allocate bandwidth for data exchange and the bidder is connected to only one ISP, such an arrangement is less efficient.

The new bidding entity 917 is preferably a bidder who entered the auction after a number of other bids have been placed. However, the new bidder 917 may be the first entity to participate in the auction. In the latter case, there is preferably a minimum bid price provided by the ISP conducting the auction.

FIG. 10 is a diagram of an Internet service provider. The ISP network bandwidth is controlled by a computer, which preferably has a central processing unit 1001, a memory 1003 connected to the CPU 1001, an input / output (
An I / O port 1005 and a data bus 1007 connected to the processor 1001, the memory 1003, and the I / O port 1005. In the example of allocating premium bandwidth, the ISP preferably performs a progressive second price auction approach. Program instructions for executing a progressive second price auction are stored in memory 1003 activated by processor 1001 to perform the techniques described in connection with FIG. ISP
Receives the bid for the auction, stores the bid in memory 1003.
Bids can also be stored in memory outside the ISP accessible by the ISP. I / O port 1005 connects to one or more networks managed by an Internet service provider.

Memory 1003 relates to the maximum amount of resources that can be allocated, the current allocation of resources and past history of allocation, and allocation processes such as payments made to ISPs after a bid has been accepted. You can also store the data. Processor 1001 may use a progressive second price auction technique to calculate the calculations required to make the assignment and the cost of the new bidder.

The present invention is not limited to the above embodiment, and many modifications and variations are possible. For example, while this application has described the solution of the problem of bandwidth and buffer space retention in a communication network, the auction technique is formalized in a manner that is general enough to be used in a wide range of situations.

[Brief description of the drawings]

FIG. 1 is a flow chart of steps describing a progressive second price auction technique.

FIG. 2 is a graphical example of resource allocation.

FIG. 3 is another graphical example of resource allocation.

FIG. 4 is a graph of a function of the utility of a bidder participating in an auction as a function of a possible bid price and a requested amount.

FIG. 5 is a graphical example of resource allocation.

FIG. 6 shows three graphs illustrating the effect of a new bidder on resource allocation.

FIG. 7 shows three graphs depicting the impact of a new bidder during an auction.

FIG. 8 shows three graphs depicting the effect of a new other bidder during an auction.

FIG. 9 illustrates an auction bidder and price system.

FIG. 10 shows a diagram of a computer processing an auction according to the invention.

──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Aurel Leather 10027 New York, NY Riverside Drive 410 New York

Claims (42)

[Claims] 1. A method of allocating resources, comprising: retrieving data representing at least one bid comprising a quantity data component and a price data component; and representing a new bid comprising a quantity component and a price component. Retrieving the resource for the at least one bid having a higher price component than the price component of the new bid; and responsive to the initial allocation step, Allocating the resource to a new bid; calculating the cost of the new bid in response to the price component of at least one bid having a lower price component than the price component of the new bid; Storing cost data representing the cost. 2. The method of claim 1, wherein said resource is a bandwidth. 3. The method of claim 1, wherein the resource allocation of the new bid and the calculated cost are transmitted to an entity that is the new bid. 4. The method of claim 1, wherein the allocated resources are utilized by an entity associated with the bid. 5. The method of claim 1, further comprising: ordering the retrieved at least one bid by a price component of the at least one bid. 6. The method of claim 1, further comprising the step of retrieving data representing a maximum capacity of the resource, wherein the allocating step is responsive to the maximum capacity.
The described method. 7. The method of claim 6, wherein the resource is allocated to the new bid only if the initial allocation is less than the maximum capacity. 8. The method of claim 1, wherein the initial allocation of resources is performed before processing the new bid. 9. The method according to claim 1, wherein calculating the cost of the new bid is responsive to the retrieved at least one bid having data representing a lowest price component allocated in the first allocation. 8. The method according to 8. 10. The method of claim 8, wherein calculating the cost of the new bid is responsive to the retrieved plurality of bids having data representing the price lower than the new bid. 11. The method of claim 11, wherein calculating the cost of the new bid is responsive to the at least one retrieved bid having data representing a lowest price received an assignment if the new bid is not retrieved. The method of claim 1, wherein 12. The method of claim 11, wherein calculating the cost of the new bid is responsive to a plurality of the received bids having data representing a price lower than the new bid. 13. The method of claim 1, wherein one of the at least one bid is made by a processor performing the assigning step. 14. The method of claim 1, wherein the at least one bid comprises valuation data having a plurality of price components associated with a plurality of quantity components. 15. The method of claim 14, wherein the assignment is responsive to the plurality of price components and the quantity components. 16. The method of claim 1, wherein the new bid comprises valuation data having a plurality of price components associated with a plurality of quantity components. 17. The method of claim 16, wherein the assignment is responsive to a plurality of price components and the quantity component of the new bid. 18. An apparatus for allocating resources, comprising: means for retrieving at least one bid comprising a quantity data component and a price data component; means for retrieving a new bid comprising a quantity component and a price component. Means for initially allocating the resource to the at least one bid having a higher price component than the new bid; means for allocating the resource to the new bid in response to an initial allocation step; Means for calculating a cost of the new bid in response to the price component of at least one bid having a lower price than the new bid. 19. The apparatus according to claim 18, wherein said resource is a bandwidth. 20. The apparatus of claim 19, further comprising means for storing data representing the cost. 21. The apparatus of claim 18, further comprising means for ordering the at least one retrieved bid by a price component of the bid. 22. The apparatus according to claim 1, further comprising means for retrieving a maximum capacity of the resource and allocating the resource in response to the maximum capacity.
An apparatus according to claim 8. 23. The apparatus of claim 22, wherein the resource is allocated to the new bid by the allocation means only if the initial allocation is less than the maximum capacity. 24. The apparatus of claim 18, wherein the initial allocation of the resources is completed by the initial allocation means before the new bid is processed. 25. The calculation of the cost of the second bid by the assigning means responsive to the retrieved at least one bid having a lowest price assigned during the initial assignment. 25. The apparatus of claim 24. 26. The apparatus of claim 25, wherein calculating the cost of the new bid by the assigning means is responsive to a plurality of received bids having a lower price than the new bid. 27. The calculation of the cost of the new bid by the allocating means responsive to the retrieved at least one bid having the lowest price that received an allocation if the new bid was not made. 19. The device according to claim 18, wherein 28. The apparatus of claim 27, wherein the calculating of the cost of the new bid by the assigning means is responsive to a plurality of received bids having a lower price than the new bid. 29. The apparatus of claim 18, wherein one of the at least one bid is placed by an operator of the apparatus. 30. The apparatus of claim 18, wherein the at least one bid comprises valuation data having a plurality of price components associated with a plurality of quantity components. 31. The apparatus of claim 30, wherein the assignment by the assigning means is responsive to the plurality of price components and the quantity components. 32. The apparatus of claim 18, wherein the new bid comprises valuation data having a plurality of price components associated with a plurality of quantity components. 33. The apparatus of claim 32, wherein the assignment by the assigning means is responsive to a plurality of price components and the quantity components of the new bid. 34. A system for allocating resources, comprising: at least one bidding entity for bidding with a quantity component and a price component; a new bidding entity for bidding at least one bid with a quantity component and a price component; Initially allocating the resource to the at least one bid having a higher price component than a second bid and allocating the resource to the new bid in response to a first allocating step, the price being lower than the new bid A processor responsive to the price component of the at least one bid to calculate the new bid cost. 35. The system of claim 34, wherein said resource is a bandwidth. 36. The system of claim 34, wherein the processor sends information representing a price component of the new bidding entity to the at least one bidding entity. 37. The processor retrieves data representing a maximum capacity of the resource and allocates the resource in response to the maximum capacity, wherein the processor allocates the resource with the first allocation greater than the maximum capacity. 35. The system of claim 34, wherein the new bid is only assigned if it is smaller. 38. The system of claim 34, wherein calculating the cost of the new bid is responsive to at least one retrieved bid having a lowest price component assigned in the first assignment. 39. The system of claim 34, wherein calculating the cost of the new bid is responsive to a plurality of received bids having lower prices than the new bid. 40. The method of claim 34, wherein calculating the cost of the new bid is responsive to at least one retrieved bid having the lowest price that received an assignment if the new bid was not made. The described system. 41. The system of claim 34, wherein said at least one bid comprises a plurality of prices associated with a plurality of quantity components. 42. The system of claim 41, wherein said assignment is responsive to said plurality of price components and said quantity component.