The purpose of the benchmark is to determine the performance metric of a stream processing system: the maximum scale at which the system can respond to the specified set of continuous and historical queries while meeting their response time and accuracy requirements. It is assumed that the benchmark will run with increasingly larger scale factors until one is found for which the requirements cannot be met.
Linear Road tests Stream Data Management Systems (SDMS) by measuring how many expressways a SDMS can support by giving accurate and timely results to four types of queries that fit two categories: continuous and historical.
To run the benchmark MITSIM, available in the download area, generates data for a selected number of expressways. This data is for a 2 hour span of vehicles travelling on an expressway(s). The stream data generated by the simulator consists of four types of tuples (queries): Position Reports and historical query requests for Account Balances, Daily Expenditures and Travel Time Estimation.
Toll Notification, triggered by a position report, tells the vehicle of the toll it will be accessed for entering the next segment of the expressway it is traveling on.
Accident Notification, an accident occurs when two vehicles are “stopped” at the same position at the same time. A vehicle is stopped when it reports the same position in 4 consecutive position reports. Once an accident occurs in a given segment, traf- fic proceeds in that segment at a reduced speed determined by the traffic spacing model. The accident takes anywhere from 10-20 minutes to be cleared once it is detected. Accidents effect the amount tolled to a car.
Account Balance Query is an account request for a vehicle. It returns the total amount in an account for any driver that requests it.
A Daily Expenditures Query is a request for a vehicle’s total tolls on a specific expressway, on a specified day in the previous ten weeks.
Travel Time Estimation is a request for an estimated toll and travel time for a journey on a given expressway, day of the week and time of day.
The Queries can be devided into two types of catagories:
Continuous Queries are queries whose results are constantly updated based on the contents of the data. In Linear Road, toll notification is based on a continuous query.
Historical Queries ask for information based on previous events stored in the database. An example of a historical query is one asking for a total account balance at a particular moment. In Linear Road, account balance requests, daily expenditures and travel time estimation are examples of historical queries.