Access to Simulations

Gio Wiederhold
Stanford University, Computer Science Department 

We recognize a need to complement passive information systems with 
active, computational services that can take data and models that define 
the past and project them forward. 

Decision-making does not only require access to passive, historical 
information, as obtained now from databases and digital libraries, but that 
information must be actively processed to allow assessment of effects on the 
future for alternate decisions. 

If we consider a timeline we find a range of services: 

...... past .............. now ....... many possible futures - libraries ---- databases -- 
interaction ------- simulation 

Simulations can be as simple as spreadsheets, as mathematical as the efforts 
of the 'Club of Rome,' as detailed as projections published in econometric 
journals, and as exciting and controversial as those published in business 
magazines. 

However, they have always been disjointed from the information systems on 
the internet. Even when their results are published on the web there is no 
interaction, so that a decision-maker or evaluator has no control over 
parameters that determine alternative actions or alternative assumptions.

Many technologies could be a vehicle to bring such futuristic services to a 
wider audience. An approach we have experimented with is the 
development of a language which extends database concepts: SimQL. 

We believe that a language which can make simulations as accessible as 
other programming components, as databases and graphical user interfaces 
(GUIs) are now, will inspire added-value vendors on the net to start making 
computational tools that provide projection into the future available.

It needs to be emphasized that a SimQL is not intended to write simulations, 
just like SQL is not a language to write a new database system. Just as SQL 
gives applications access to a wealth of database technology and to database 
services that are often maintained by others, we expect that a SimQL will 
provide access to simulation technology and simulation services maintained 
by others, and break the bottleneck which is now experienced when 
simulations are to be a part of larger systems in planning for industry, 
military, and business. 

A simple illustration of our intent can be gleaned from showing two queries, 
the first one addressed to a database and the second one to a simulation:

1. SELECT Temperature, Cloudcover, Windspeed, Winddirection FROM 
WeatherDB WHERE Date = `yesterday' AND Location = `ORD.' 1. 
PREDICT Temperature, Cloudcover, Windspeed, Winddirection FROM 
WeatherSimulation WHERE Date = `tomorrow' AND Location = `ORD.' 

The models that drive simulations will be of necessity more complex than 
database schemas, and one objective of the research would be to see how 
little need be exposed to the user's application programs. A SimQL will have 
to address some issues that do not arise when SQL deals with databases, 
because simulations have features that differ from databases.

Specifically, any projection into the future will be associated with an 
uncertainty that must be reported. A decision-maker can then have the 
choice of moving to a future with lower benefits and less risk, or higher 
returns and more risk. Also, the systems that employ simulations must be 
able to handle and weigh multiple futures, whereas we assume that there is 
only one correct past. 

The objective of this description is not to advocate a specific system or 
language, but to illustrate a direction that is not available in an 
integratable fashion from modern information systems. 

Making information systems active could be a great step forward for the 
Next Generation Internet, and exploit its computational as well as its 
communication resources.