3. Experimental Analysis

We have implemented the two new approaches and tested them on a sample NASA data set. The code was written in Java Programming Language and the tests were run on a Local Area Network. A SUN SPARC ULTRA 1 workstation was used as the server and a Pentium P-166 Processor based PC was used as the client. The Global Change Master Directory Query Preview Panel was used as the basis for the Java Code. The sample NASA data set, with 60275 tuples, was obtained from the NASA V0 Client Homepage. We have used longitude, latitude, and year as our preview attributes. We have also modified the previous code to download this sample data set and incorporated results of this version to our report to show how previous approaches perform with this sample data set in comparison with the two new approaches. Figure 7 shows a snapshot of the Query Preview Panel used in our tests and Figure 6 gives the comparative results of these tests.

Previous

Approaches

Post-Fetch

Approach

Pre-Fetch

Approach

Querying Time

2-3 seconds

1-2 seconds

0-1 seconds

Client

Main-Memory

Utilization

750 Kbytes

None

50 Kbytes

Setup Time

25-30 seconds

None

10-15 seconds

Server

Disk

Utilization

750 Kbytes

200 Mbytes

750 + 50 Kbytes

= 800 Kbytes

Data

Preparation

Time

None

Within 1 day

Within 1 hour

Restrictivity

Level of the

Preview Panel

None

None

1 selection

per

attribute

Figure 6: Experimental analysis of the approaches (previous, post-fetch, and pre-fetch).

 

Figure 7: The Query Preview Panel used in our experiments.