Economic Framework for Pricing and Charging in Digital Libraries
Computer Science Department
New York, NY, 10027
University of Crete and ICS-FORTH
GR 711 10
D. F. Ferguson
IBM T.J. Watson Research Center
Hawthorne, New York, NY, 10532
Computer Science Department
New York, NY, 10027
D-Lib Magazine, February 1996
We consider commercial Digital Libraries as information economies
consisting of several players: authors and publishers who create and sell their
collections, suppliers (e.g. computer systems) who provide information storage,
indexing and access services, information-agents who provide searching and
presentation services, and users who request for services. In such an economic
framework, one can envision suppliers and information-agents competing to
provide services for information storage, searching, access and presentation.
In providing such services, several issues arise; among them are socio-economic
and cultural aspects of pricing information objects and Quality of Service
(QoS) to access and view these objects. These issues play an important role in
allocating resources such as processing time, network bandwidth and buffers,
memory, cache and I/O, which are distributed (and owned) among various players
in the economy. Using this framework, we present the interactions among the
players, service models, pricing and charging/billing mechanisms, and
corresponding implementation issues in large digital libraries.
Digital Libraries will have a major influence on the design of future
information systems. They are the cradle from which future advanced information
technologies will emerge to provide "transparent" services to a variety of
users. Digital Libraries have already attracted many real-world agents, which
participate for various economic reasons. For example, publishers of books are
in the process of presenting their collections in digital formats in order to
reduce production costs and increase profit margins.
Digital Libraries will house information objects in various media, such as
text, audio, video and image, and provide information access services to a
variety of users. One can envision a plethora of applications, such as
collaborative computing and virtual class-rooms, using such services. Users of
these applications, being heterogenous in their preferences, could request for
different qualities of service for information access. The preferences of users
could be based on the charge to access information.
For example, users browsing through a Digital Library of video clips or image
clips (as shown in Figure 1) would prefer different qualities of service based
on the charge and the availability of local resources (local software and
hardware). In the Figure 1, we also illustrate that fact that current networks
have limited bandwidth, and efficient usage of bandwidth is neccessary to
support many concurrent accesses to the Digital Libraries. Therefore, one can
envision several levels of services being offered by Digital Libraries to
access information objects. In the example, we show that a video object can be
presented to the user in different formats (MPEG, MPEG-II, etc.).
In designing service mechanisms for accessing information objects in digital
libraries, several issues arise. One of them is the issue of pricing services
(and multiple service levels) to access information objects, and accounting and
billing users for QoS. These are economic issues, however, they play an
important role in allocating resources such as processing (CPU) time, network
bandwidth, buffer, memory and cache for storing, searching, accessing,
transporting, and presenting various information objects in a coherent way to
the users. For example, users might want to know the prices for different
qualities of service to access an information object, such as a video clip of a
lecture based on a specific day (and time).
In order to develop a framework for pricing services and charging/billing users
for access, several questions arise and some of them are as follows: What is
the architecture? What is the economic framework? Who are the players in the
economy, and what are their interactions? What should be priced? What are the
pricing models? What are the practical issues in implementing pricing schemes,
and charging frameworks in large distributed Digital Libraries? There are
several accompanying system level issues for storage and access of
objects. Some of them are as follows: What are performance demands on the local
operating systems and distributed system as a whole? What are appropriate
scheduling and load-balancing mechanisms? What are the object storage and
Economic models for storage and replication of information objects have been
investigated by Ferguson
et. al and
Stonebraker et. al.. Economic models for resource management, in general,
have been considered in  and 
. Considerable research in the area of Information economies is being done in
. Also, in , a collection of
references to related works in this area is given. Work in the area of
autonomous, profit-seeking agents in market economies is being conducted in .
Current networks and systems have limited resources (such as link bandwidth).
Therefore, allocation of resources, such as processing time and network
bandwidth, can be crucial in providing efficient services. In the future,
bandwidth and processing time will be more than sufficient, and information
objects will be priced based on the demand for them rather than on the Digital
Library system (DLSystem) that houses them. We consider a DLSystem to be a
complex computer system that provides storage, indexing, and access services to
various information objects. In this paper, we assume that a set of resource
allocations make up a service. An information object could also be a resource,
but we do not price information objects in our current models, however, we plan
to incorporate this in the near future.
Our main approach in pricing and accounting for QoS is an economic model of the
players in large Digital Libraries. This involves modeling and understanding
the various tasks that are required to support Digital Libraries. There are
several reasons why this form of modeling helps in understanding the interfaces
between the various elements of Digital Libraries. The main reasons are as
The novelty introduced by the economic approach is not only in
modeling large decentralized, autonomous systems, but also in designing
efficient systems that account naturally for user QoS requirements, and
optimally allocate resources for the various services in Digital Libraries.
- Limiting Complexity: Economic models provide several interesting
contributions to resource sharing algorithms. The first is a set of tools for
limiting the complexity by decentralizing the control of resources. The second
is a set of mathematical models that can yield several new insights into
resource sharing problems (such as scheduling in computer systems, object
replication in distributed systems).
- Decentralization: In an economy, decentralization is provided by
the fact that economic-agents attempt to achieve their goals in competitive
fashion (selfish fashion). There are two types of such agents,
suppliers and consumers. A consumer attempts to maximize its
benefit by obtaining the services or resource allocations that maximize its
preferences under a wealth constraint (see ).
Suppliers compete in order to maximize revenue and minimize costs, which means
attract more users and provide the best services.
- Pricing and Coordination: Most economic models introduce money
and pricing as techniques for coordinating the behavior of agents. Each
consumer is endowed with money, which they use to purchase services or
resources (see ). The price a supplier charges for a
resource or a service is determined by its supply and the market demand.
- Usage Accounting, Billing and Dimensioning: In using economic
models for service provisioning in distributed systems, usage accounting
becomes a part of the economy. Suppliers have to keep track of resource usage
in order to price resources effectively, and thereby charge/bill the users
- Administrative Domains: Large distributed systems and computer
networks are spread over several domains. Each domain is typically managed and
controlled by an administrator. An economic model of such an environment would
imply that administrators advertise their services and prices. Therefore,
providing a simple interface to negotiate between domains.
- Scalability; A key issue in designing architectures for
services in large computer networks and distributed systems is
scalability. Models of competition provide--in a natural
fashion--mechanisms for scaling services appropriately based on the
service-demand and resource-availability.
Simple Model of Digital Libraries
We consider three kinds of players in the economic model of Digital Libraries:
Suppliers (storage and indexing), Agents (who search, transform and present
information) and Users. The three players are described below:
In Figure 2, we illustrate the economic model with three players. There are
many classes (K) of users who obtain services from one or many agents. The
agents buy services from one or more of the DLSystems (Library Systems), and
provide searching and presentation services to the users for a fee. In the
Figure, users are shown to get services only through the agents, however,
one can also model users obtaining services directly from the DLSystems, which
means that users have to search, on their own, for information about Digital
- Suppliers: Commercial DLSystems buy works from various publishers
and authors, and provide services to a variety of users for a fee. The
DLSystems not only price information but also price QoS to access the
information. Pricing is based on the user demand, and users represent their
demands through agents. The prices remain fixed every time-interval, assuming
that time is divided into intervals. For example, a time-interval could be a
6 hours in a day. The time-intervals are decided by each DLSystem
independently based on the market demand. The suppliers provide storage, access
and local indexing services.
- Information Agents: These are suppliers that provide
Value-Added Services, such as transparent searching and presentation of
objects to the users. From now on, we refer to Information Agents as
agents. The agents charge users for QoS provided. Agents contract services and
resources from DLSystems over a period of time, and provide stable set of
services to the users. The stability is provided by a flat service fee over
period of time (example: half a day) to the users. Agents renegotiate for
services from each DLSystem provided it is profitable, else they release
resources. The agents compete to provide One-Stop Shopping services
to the users.
- Users: They present complex queries to the information agents, and
request for a service in viewing or obtaining the objects. Users could be
grouped into query classes based on similar access patterns or just based on
service required. Users within a group or class share the costs of the
services. Users are free to choose among the various agents based on the QoS
Pricing Services: Searching and Presentation
In the economic framework, agents compete for resources and services from
DLsystems. They buy information access services ahead of time, and provide a
set of integrated services to consumers (users). Similarly, DLsystems which
have common information objects naturally compete to attract users and maximize
profit. In designing pricing schemes for such services in an economic
framework, several considerations need to be taken into account. The main
considerations are the following:
Agents, post their prices and services offered for all the time-intervals on a
certain day. For example, agents might offer low prices for a service between
6.00 A.M. and 4.00 P.M. and offer a higher price for the same service between
4.00 P.M. and 10.00 P.M. Agents, based on factors such as user demand profile
and market behaviour, might compete to buy more services from the DLSystems.
- What should be priced? Resources (such as buffer, bandwidth and CPU time)
and services, based on performance parameters such as response time should be
priced. Services for quality of search can also be priced. Information
objects, in general, will be priced, and we plan to study this in the future.
- Prices must be set such that demand equals the supply. This is done to
make sure that resources are not over-booked, and service quality is always
- Prices should reflect user query-request behaviour (load). This is
essentialy to control congestion of requests to a single digital library
- Prices should not fluctuate rapidly as users can get confused. In
practice, prices should be stable for a reasonably long period of time.
- Prices need to be negotiated ahead of time. This means that arriving users
choose the services from agents based on the price information and the service
Agents have bugdets for buying services from DLsystems. Agents compete among
themselves for services from suppliers. Agents charge the users based on QoS
requested in searching and presentation. Suppliers price resources/services
based on demand from the users. We assume that time-intervals, where prices are
fixed, are set independently by each supplier.
For example, supplier A might have 6.00 A.M till noon as one time period, and
from noon till 6.00 P.M. another time period and price structure, and from 6.00
P.M. till 10. P.M. a different set of prices. Supplier B might have fixed
prices between 6.00 A.M. and 5.00 P.M and another price structure for 5.00 P.M
to 11.00 P.M. The behaviour of the suppliers is an economic one, the change in
prices is due to the market demand.
The suppliers operate indepedently in selecting the time periods for stable
prices, however, they are driven based on the collective market
demand. Similarly, each agent independently selects the time periods for stable
prices to the users.
Price Negotiation: The suppliers, based on the demand profiles of
the agents, decide on times when negotiation of the services and prices is
done. For example, the price negotiation times everyday could be 6.00 A.M and
4.00 P.M. for supplier A, and 6.00 A.M and 5.00 P.M. for supplier
B. Negotiation of prices and services by the suppliers is done at the beginning
of every time period. Prices could be announced just after the negotiation.
In economic terms, the negotiation protocol between the agents and the
DLsystems could be done using the tatonnement process: agents are
endowed with some wealth. Each agent computes the demand from a benefit
function and wealth. The benefit function is simply a utility function as
described in Ferguson
et. al.. The aggregate demand from all the agents is collected by each
supplier who then computes the new resource price. If the demand for a
resource is greater than its supply, the supplier raises the price of the
resource. If there is surplus supply, price is decreased. With the new prices,
agents again compute and present their demands to the suppliers. This process
continues iteratively till the equilibrium price is achieved where demand
equals the supply.
Bidding and auctioning for resources/services are other forms of price
fixing. There are several auctioning mechanisms such as the Sealed Bid Auction, Dutch
Auction, and English Auction . The basic philosophy behind auctions and
bidding is that the highest bidder always gets the resources, and the current
price for a resource is determined by the bid prices.
We model the digital library system as a simple computer system with CPU and
storage (disks, memory and cache). We assume that storage is very large and
therefore the charge is negligible, however, we assume that processing power
and network I/O bandwidth is limited, and users are charged for them. Users who
wish to get a good response time, pay for network I/O bandwidth and processing
Some simple performance based economic models are discussed in . Similarly, models for allocating network bandwidth and
buffers for various services are given in , , , 
and . Our goal is not only to investigate pricing
mechanisms, but also design architectures to support various pricing policies
and charge users for QoS provided. More details on the modeling and pricing,
and performance models (e.g. queueing models) for QoS provisioning (such as
response time and video/image quality) is given in .
- Long-Term Pricing: One can argue that prices should be set on a
monthly or weekly basis. However, this can cause potential problems as multiple
levels of services cannot be supported flexibly for the various users. For
example, users may not want services on all days, instead they would prefer to
get services on demand.
- Trade-offs: If the agents in the information economy do not exist,
then users will have direct access to the DLSystems. In such situations, users
will send requests for information at arbitrary points in time. Therefore,
guarantees for services and prices can be complex, as several thousands of
requests could potentially arrive. If prices are not stable for long periods
of time, then users can get confused about choosing the right agent.
Charging, Accounting and Payment
Users will see a charge for access ahead of time, and will have a choice in
choosing the right agent for searching and access services to the information
objects. Assuming that prices fluctuate slowly, several practical issues of
billing and payment can be resolved very easily. There are already a variety of
payment models (see ), and a whole new chapter has begun
in this area with the multi-industry project initiated by CommerceNet and the W3 Consortium.
Implementation Issues and Platforms.
We are currently investigating ways to implement a charging/accounting
framework over NCSTRL , which is an
evolving, distributed Digital Library System for Computer Science Technical
Reports. NCSTRL currently uses DIENST, which uses WWW framework (HTTP and
HTML) for communication and searching, and CGI for interaction with Web
browsers. We plan to extend DIENST to support a charging framework as shown in
Figure 3. The extensions are going to be done on a testbed of DIENST servers
that can support objects such as video clips of lectures. For example,
universities and research institutions might sell services to view video
lectures for a small fee.
In Figure 3, a distributed name service architecture along with the DIENST
architecture is shown. The name service is designed to provide information
about services offered by the Digital Library servers (DIENST servers). This
information includes service prices and copyright information to access the
objects. In the Figure, we also show a Digital Library site (example: campus or
an organization), where several WWW based clients query for information from
the Digital Library Server.
Our goal is to give to the user, along with the search/query results, the
qualities of service offered and the corresponding price for access. For
example, if the objects are multimedia objects, then along with the search
results come the service levels (MPEG or JPEG quality video) offered by the
corresponding Digital Library servers and the price to access those objects. In
Figure 4, we show the model that we are currently building and testing. DIENST
currently returns a list of URLs when a user submits a query (keyword). We plan
to extend this by adding QoS and price information along with each URL in the
In Figure 5, we show an example of a payment scheme that we plan to study. The
payment is between a stock broker and a consumer, and between the Digital
library system and the consumer. The Digital Library contains information
about companies (their earnings, stock performance, and other related
information) and news papers. The user first queries the Digital Library for
stock information such as stock price, earnings and performance plots of a
specific company or a collection of companies. The user then decides on stocks
to invest and contacts a stock broker to submit the request. The stock broker
buys stock for the consumer using the consumer-account in the bank. The
interesting issues are in storing and presenting information about companies in
Digital Libraries and payment mechanisms.
In this story, we presented an economic framework for pricing services and
charging/billing users in Digital Libraries. In this framework, we presented a
competitive model of interaction among the players of the economy. Our goal is
to use such a framework to support various other pricing schemes in order to
charge/bill users for services. We also presented an architecture for
charging/billing in a Digital Library system using the DIENST technology. Our
future work is to study and implement various payment schemes in large Digital
Libaries, and to investigate issues of copyright and protection. We also plan
to investigate issues of caching and prefetching in large Digital Libraries,
where agents sell caching services to consumers.
The research work in this paper were supported in part by the Center For
Telecommunications Research, Columbia University, New York, in part by the
LYDIA project (ESPRIT III P8144), and in part by the CASHMAN project (ACTS AC
-  D. F. Ferguson,
C. Nikolaou, J. Sairamesh, and Y. Yemini, Economic Models for Allocating
Resources in Computer Systems, Market based Control of Distributed
Systems, Ed. Scott Clearwater, World Scientific Press, 1995.
-  J. Sairamesh, D. Ferguson, and Y. Yemini,
An Approach to Pricing, Optimal Allocations, and Quality of
Service Provisioning in High-Speed Networks,
Proceedings of the INFOCOM'95.
-  D.F. Ferguson, C. Nikolaou and Y. Yemini,
``An Economy for Flow
Control in Computer Networks,'' Proc. of the INFOCOM'90.
-  D.F. Ferguson, Y. Yemini and C. NiKolaou,
Algorithms for Load Balancing in Distributed Computer Systems,''
Proc. of International Conference on Distributed Systems (ICDCS'88).
Service Models and Pricing
InterPay: Managing Multiple Payment
Mechanisms in Digital
by S. Cousins, S. Ketchpel, A. Paepcke, H. Garcia-Molina, S. Hassan, and M.
to appear in Digital Libraries 95
Connection Establishment and Pricing
The Information Economy
-  J. Sairamesh, C. Nikolaou, D. F. Ferguson, Y. Yemini,
Pricing Services in Digital Libraries , Working paper, to be
published as a joint Columbia University CS-TR, New York and
ICS-FORTH CS-TR, Crete, Greece, 1996.
-  C. A. Waldspurger, T. Hogg, B. Huberman, J. Kephart,
and S. Stor netta, Spawn: A distributed computational ecology, IEEE
Trans. on Software Engg., 18(2), February, 1992.
Copyright © 1996J. Sairamesh, C. Nikolaou, D. F. Ferguson, Y. Yemini