The WALRAS Blue-Skies Economy

A simple illustration of the market-oriented programming approach to allocating information services over a network.

Designed and implemented by Tracy Mullen and Michael Wellman.

WALRAS formulation:

Overview

The Blue-Skies economy was designed to experiment with a simple model of information service provision on a network. In this model, a popular information service (of which the canonical example is Blue-Skies, a weather-information server based at the University of Michigan) is available on the internet, and local agents decide whether to serve as mirror sites for the service.

This economy was developed as an extension to the WALRAS generic transportation economy. Our description below describes the differences.

Goods

In addition to transport on a link and basic transportation resources (in this case, effective network bandwidth), the available goods include:
  1. Information goods at a location. For example, Blue-Skies@local-site1, representing an amount of information service available at a particular node in the network. There is one of these goods for each (service,node) pair. Amount of the good is a combined measure of quantity and quality of service.
  2. Basic computational resources. The various resources needed to provide caching service: storage, processing, plus another good that is an amalgam of the two.

Agents

End User Consumers

Consumers in the Blue-Skies economy are end users of information services. As in the transport economy, the consumers trade generic resources (network and computational) for the good they care about--Blue-Skies at their local site. Unlike shippers in the transportation economy, the Blue-Skies consumers are not given an absolute quantity requirement, but rather have preferences for various combinations of resources and service levels. By specifying a positive preference for retaining some of their endowed resources, we allow that the consumers have some other (implicit) uses for the resource.

The consumer agent bids to maximize utility subject to its budget constraint, in the conventional manner.

Producers

In addition to carriers and network arbitrageurs, the Blue-Skies economy includes three other producer types. Manufacturers produce particular computational resources (e.g., processing, storage) or information goods (Blue-Skies) from the generic (amalgamated) resource goods. Delivery arbitrageurs are similar to transportation network arbitrageurs, except they bundle information services at a location (e.g., Blue-Skies@internet) with transport from that location to another (e.g, local-site), to produce that service at the second location (Blue-Skies@local-site). Finally, the mirror provider has the capability of transforming local storage and other resources into provision of the information service at its local site. It can also choose to access the service from another site instead of mirroring, in which case it acts just like the delivery arbitrageur.

Results

Depending on the initial configuration, the Blue-Skies economy exhibits a range of interesting behaviors. If there is sufficient demand at a local site, the mirror provider will set up a local cache. If there is not sufficient demand locally, the mirror site may still be profitable if the service can be resold to other sites.

For some values of the parameters, the behavior oscillates. When there are no mirror sites, there is a lot of network traffic and a relatively high price of bandwidth, and thus it appears profitable to set up a mirror site. But once the mirror is set up, traffic decreases as does the price of bandwidth, and so direct access to the internet appears to be cheaper. This phenomenon is a direct consequence of applying the competitiveness assumption when there are a small number of agents. With a larger number (bigger network, more diverse set of services available), the effect of one mirroring decision would have a negligible impact on overall traffic, and hence this oscillation would be ameliorated.

Extensions

In ongoing and future work, we are extending the Blue-Skies economy in several ways:

References

We have a paper describing this work:

T. Mullen and M.P. Wellman. A simple computational market for network information services. to appear in First International Conference on Multiagent Systems, June 1995.

See the references in this paper for other related work.


Updated by Michael Wellman on 4/21/95