Abstract
This experiment was run following a visit to the University of Queensland, where a simulation engine has been created using MatLab SimuLink. This simulator is developed for use as the underlying engine in future User Interface (UI) experiments. The simulator engine currently gets/puts information via a DDL link to an MS Excel spreadsheet. While this is suitable for initial testing, when UI experiments are run the data transfer method will need to be faster. There is a need for more data to be collected in order to accurately assess system state at any time during the experiment.
There are two methods currently proposed to achieve this latter goal. One is to send a record to the database when the state of some component changes. The other is to capture the entire system state at pre-determined times then save the entire system state. The former method is expected to have less impact on resources, but prevents making temporal comparisons (unless you wish to replay the entire experiment). If possible, the latter method is preferred.
It was recognised that some bottlenecks exist under the current system design. Notably, the simulation engine requires significant processor power to run effectively and that the resources required by MS Excel slow processes. The experiment has the following goals:
Is a single, large, transaction more efficient than multiple smaller transactions?
Does machine speed/specifications have an effect on how the experiment runs?
Does the ODBC processing overhead have an adverse impact on data transfer speed?
Does network connectivity speed have an impact on data transfer speed?
Does the number of fields being transferred have an impact on data transfer speed?
By identifying what combination is most effective we can then determine an optimal hardware setting for future UI experiments.
This experiment is comparative in nature. It is not fully robust due to limitations in tool availability. However, the comparisons made may be useful in eliminating some of the possibilities and guiding further experiments.
Three hardware setups were used. Two computers are similar in specification, but with different network connection speeds. The third was a much lower specified, but served to provide a comparative role – how much difference does improved machine specification make?