5. Limitations and Further Work

There are still some limitations which affect the usefulness of NetSpy, mostly to do with the user interface.

When constructing a scene, it is not possible to undo a step. For example, if the user places a Cube gadget into the scene, but then decides they would actually like to use the StretchyText gadget instead, they must start constructing the scene from scratch. While this does not limit the functionality of the system it can make it quite frustrating to use.

Once a scene has been created and is being used to monitor statistics, only limited user interaction possible. It would be quite useful to see the actual value of a statistic rather than just the representation provided by the gadget. This could be done by allowing the user to select a gadget and then displaying in text somewhere in the display the raw values of the statistics that are affecting it.

The layout of gadgets within a scene is very simple minded in the current implementation, which may cause the gadgets to be widely spaced. This means that it may only be possible to make out see details of a small number of gadgets at once. By improving the layout algorithms this problem could be rectified. I would like to point out that this is not a design limitation. My design allows for easy introduction of improved layout algorithms (Arranger gadgets), however the work of implementing them has not been done.

Another feature which I had designed, but time constraints prevented me from implementing was that of hierarchies. A hierarchy is a special kind of distributor gadget with two distinct visual representations. One involves using one copy of the chosen shape gadget to view the statistics, no matter how many instances there are. The values chosen to update the gadget would be an aggregation of all of the instances. The second visual form would be the same as for a normal distributor. That is, one copy of the gadget for each instance of the statistic. This would allow the scene to be less cluttered, but the details about each individual instance would be available if desired.

Silicon Graphics' PMCS was used as the basis for the collection of statistics from the various hosts as well as the design of DPML-MCS. However, the unavailability of computers running the software meant that the final step of integrating support into NetSpy was not completed. Enough testing was done to show that this was indeed possible, and some code has been written to demonstrate the feasibility of this. The fact that DPML-MCS is based on PMCS means that the mapping of values from one to the other is simple.


[Table of Contents] [Previous Section] [Next Section]


David Sykes / dsykes@cs.monash.edu.au