Grid Markets
From EStar
Grid Markets Project
Members of the eSTAR team are partners in the Computational Markets Project. The project attempts to answer two questions;
- Can the emerging grid and web service technologies that are being developed be used to support markets in computer based services in astronomy?
- What parameters can the particpants in the market negotiate on in an astronomical context?
e-Markets for Astronomy
The establishment of an e-market for the exchange of telescope time is perhaps one of the few scalable ways of providing access to the growing number of robotic and remote telescopes, and architecturally perhaps the easiest way of allowing autonomous real-time access to these telescopes in software, without having to make number bi-lateral agreements between the existing and future networks.
While there are numerous ways in which this potential e-market could be established and operate, these can generally be broken down into three models; a centralised market, a semi-centralised market and a distributed peer-to-peer market. The main question outstanding in all these models is the currency used in the market.
Centralised markets
In the centralised model, a central server keeps track of the time allocated to each "trader". Both time held by them, and available for them to use, and time which they are offering for "sale" onto the market. Here each trader sells telescope time into the market, and currency is added to their account when another trader buys time from time. The currency can be used purchase telescope time from other observatories, with the time bought and sold being tracked centrally by the market. In a truly centralised market observation requests and data return is also passed via the central server, and not directly to the individuals.
Semi-centralised markets
A centralised market has one immediately obvious problem, it represents a single point of failure for the entire network. So it is important to realise that what is traded is the time itself, the e-market does not necessarily have to directly drive the telescope. The market need only know how much time is for sale, and the price at which it is to be sold. In this model access to the telescopes would still therefore be via their own embedded software. However to avoid centralisation the market should not keep track of time sold, or at least not be the sole source of this information.
There are at least two obvious ways this could be implemented. Either the market could inform the telescope as to the identity of the user buying time, and book keeping as to who is owed what time is done at the telescope itself, or the market passes the user an encrypted token for a specific telescope, or network of telescopes. This token can be then be passed (redeemed) by the user with a specific telescope, or with a group of telescopes, or possibly with any telescope.
Both of these avoid the obvious choke point, the market, when we want to make an observation request. Hence, once the transaction is completed to purchase "time", the market itself is no longer essential. This is an essential feature and will allow observing to continue, at least for a time, if the market itself is suffers an service outage or other disruption.
Distributed markets
The final model is a distributed market, and most closely resembles the real economy. Here the market is distributed, and operates in a peer-to-peer manner, with negotiations taking place between a number of different "traders" to settle on a price for an observation. Here digital tokens are again used, however unlike the semi-centralised market this model is more closely analogous to the "black economy" without any central oversight, the tokens themselves serving as a digital currency between the traders. However while there is no centralised, or even a semi-centralised market, ad-hoc markets may be formed, perhaps around specific observing programmes, by the autonomous software trading in the time for those observations. While perhaps the most efficient, it relies heavily on the emerging complexities of an agent architecture model\cite{Allan06a} and offers little scope for direct human intervention. It would be difficult for a human to be a direct participant on such a distributed framework.
The stone age market
The first steps towards a market based economy in telescope time is the so called "stone age market". Here a limited scale prototype, with only a few participants, will test the free market approach to scheduling. By constraining the number of projects involved the sociological problems can be avoided at least in part by drawing up a number of bi-lateral agreements between the participants for the exchange of telescope time. This limited market will serve both as an architectural test bed and a sociological proof of concept for the free market approach. Implementation details for the stone age market will form one of the core topics at the second HTN workshop.
However in broad terms in a stone aged market, the participants will contribute some of their telescope time into a common robotic or service observation request pool from which the members can draw dynamically, according to how much they have contributed and participants in the market will guarantee that requests made to their telescope will be handled with high priority.
A new currency?
One of the main outstanding issue with respect to e-markets for trading telescope time, is that basis for the underlying currency?
For instance one proposal is that the telescope time itself should be the underlying currency, for instance taking a "normalised hour" as an hour of 1 m gray time on an imaging telescope (including observing overhead) with 1 arcsec seeing, scaling this unit by effective aperture and sky darkness and inversely with seeing. With spectroscopy costing more depending on a (semi-arbitary?) scaling factor and with the resolution of the spectrograph. However some arbitrary multiplier by the provider would also have to be factored into the "price" when making telescopes available on the market, simply to reflect the relative over-subscription rates of different telescopes and allow the provider to manage demand.
This semi-deterministic model has problems, the most obvious is how to "price" an unique instrument against a more generic imaging camera. One of the central requirements for a free market is the availability of multiple suppliers, which gives rise to competition that provides, without competition a market economy may not yield the efficiency gains that would be expected.
While it is still under debate, it is likely that such an approach may prove effective in the short term and allow operations of the "stone age" market. In the longer term it is possible that the only medium of exchange that will be seen as fair by all the participants in the market for telescope time may be real monetary units, of whatever currency.
