In this paper, we describe the prototypical implementation and evaluation of a database performance monitoring tool for large database management systems (DBMS). These DBMS provide the technological background for many complex e-Government applications and the availability of the managed data is crucial. The implementation focuses on the creation of application-specific gestures on a touch input device, such as the Apple iPhone 3G with software version 2.2.1. A perspective wall is used to display the data in conjunction with an acoustic indicator for navigating through the information space non-visually. We report on an exploratory investigation of the prototypical monitoring tool based on an evaluation with two groups of users: inexperienced users with no database-related professional background and users, whose daily work is closely related to database monitoring. We conclude, knowledge in the area of the respective application is helpful to make better use of the prototypical tool. Furthermore, the flexibility of the perspective wall as the visualization of choice is shown by the good overall user acceptance. Finally, the acoustic indicator gives an idea of how to support even visually impaired users in finding occurrences of problems in large information spaces, such as database performance criteria.
INTRODUCTION
The increasing growth of digital data
Since the number of people using a computer and surfing
the Internet grew from about 19.5 million in 1997 [
DBMS in e-Government applications
In order to provide centralized data management and to use
synergetic effects of shared knowledge domains, many
regional and national governments in Germany enforce the
establishment of new IT infrastructures. Examples are the
LUSD system (Lehrer und Schüler Datenbank, teacher and
pupil database [
The importance of mobile database monitoring
As the description of the two examples above suggests, the
availability of the maintained data is crucial. Therefore,
database administrators (DBAs) take care of monitoring
and optimizing the databases, keeping the systems up and
running. This task needs to be performed throughout the
day, independently from the current position of the DBA. A
failure of a database might not only result in a costly
unavailability of data, but also in a loss of sensible data,
which is inacceptable especially for e-Government
applications. Especially the complexity of current solutions
for database performance monitoring implies the need for a
desktop computer system or at least a fully featured web
browser. This contradicts the prerequisites mentioned
before and leads to the idea of creating a mobile application
with a reduced feature set, similar to the “schema later”
approach by Jagadish et al. [
RELATED WORK
As mentioned previously throughout this work, many
similar approaches in terms of visualization and support for
visually impaired people are available. Nevertheless, they
have mostly a more narrow focus of application. An
example for an application specialized on supporting blind
users in using touch-based interfaces is Slide Rule [
Similar to the tilt control of MultiStates is the
speeddependent automatic zoom approach of Eslambolchilar et
al. [
DESIGN OF MULTISTATES
To offer a mobile solution, we designed the MultiStates
prototype, which makes use of geometric and semantic
Zoom, leveraging a modified version of the perspective
wall [26]. MultiStates runs on an Apple iPhone 3G,
currently with software version 2.2.1. MultiStates provides
a new way of monitoring databases on the go by combining
two IBM products used as desktop solutions: IBM DB2
Performance Expert V3 [
Applied interface design principles
When it comes to working on a mobile platform, users need
to be supported by a tailored interface, which comprises
intuitive forms of interaction and simple visual cues.
Furthermore, the option to configure or filter the displayed
data should be given for expert users. We used Nielsen’s
[
Database state visualization
To ensure a clean and simple visualization on the screen of
the iPhone, the perspective wall has been chosen. This
selection results from a comparison of multiple
visualization techniques, which concentrate on the
presentation of a central focus region, while avoiding the
desert fog [
Interaction techniques
As mentioned previously, we tried to offer the users
redundant interaction techniques to allow the selection of a
preferred method. Most importantly, we added two new
gestures. First, shaking the phone results in resetting the
Health Summary back to its original zoom level (in case the
user lost orientation within the large dataset). Second, we
designed the tap hold/tilt gesture to allow zoom operations
single-handed. By default, the iPhone zoom operation is
executed through the pinch gesture, which makes the use of
both hands necessary. For tap hold/tilt (Figure 1) the user
simply taps and holds a finger on the display. Then, he/she
may zoom in by tilting the phone to the right or zoom out
by tilting the phone to the left. Other interaction techniques
are as shown in Table 1: swiping the finger across the
display to scroll in one of four directions. Tapping for item
selection and tilting in one of four directions to scroll. It is
to note that the double tap gesture is not used, since all
navigation tasks have been covered by the gestures defined
beforehand. Furthermore, no special drill-down gesture has
been implemented (e.g. drawing a circle around an item or
drawing the letter “L”, such as in [
Systems, such as LensBar [
Table 1. Interaction techniques available with MultiStates. category, the user may hide complete databases if no alert or exception has occurred within this category. This may also be accompanied by an audio indicator, which can be used for navigating blindly through the dataset.
Support for non-visual exploration By enabling the audio indicator and a filter category, a sound is played once an alert or exception in this category is displayed at the highest zoom level within the nondistorted central region of the perspective wall (e.g. category “Locking” of “Database 1” in Figure 2). This indicator can be used in conjunction with tilt navigation to explore the information space without needing to watch the device’s display all the time. Hence, the indicator not only lowers the need to pay attention to the running application for sighted users, but also gives an idea of how to support visually impaired and blind users. However, it is obvious that acoustic feedback may be inappropriate in silent places (e.g. libraries) or noisy environments (train stations, for instance), we decided to make use of an audible indicator for testing purposes and for power saving reasons. Using the vibration control of the iPhone for indicating problems within the Health Summary may lead to decreased battery life when the application is used every day. Even though the creation of an interface for blind users was not the focus of this work, the evaluation results provide insight into further ideas on how to better support both user groups (sighted and blind).
EVALUATION OF MULTISTATES We conducted a summative evaluation with six participants in total. Although database administrators are the intended target user group of MultiStates, it was not possible to have DBAs evaluate the system. Instead, we decided to compare three inexperienced users with no professional IT- or database management-related background (Comparison group) with three IT professionals (Expert group: IBM DB2 Performance Expert developer, tester and user experience professional). All members of both groups did not have relevant experience in using an iPhone or other touch-based devices and are sighted.
Each participant was interviewed separately, while being watched by one evaluator. The evaluation was based on two questionnaires: First, a sheet containing tasks to perform using the Health Summary and its filter options. Some tasks were timed and users had to justify why they chose a certain interaction technique while the number of errors made during interaction was counted. Second, a questionnaire focusing on each participant’s usage experience by providing scales to rate the satisfaction and acceptance of application parts (such as the perspective wall, interaction techniques or the complexity of MultiStates compared to a desktop product). Besides the questionnaires, the participants were asked to “think aloud” while they worked with the application. All comments given were transcribed and used for further analysis and interpretation of the questionnaire results.
Evaluation Findings To offer a better overview of the results of both groups, we first had a look at each group, before we directly analyzed similarities and differences of both groups. Specific results are shown for each group, while more general results are presented when both groups are compared. Numbers presented in brackets refer to the group’s mean value and the best possible value. Standard deviation is not given due to the small number of evaluation participants (except for the number of corrections needed per group and task). Comparison Group Generally speaking, the Comparison Group provided only high-level feedback. This may be due to the general lack of experience with electronic handheld devices. However, this feedback is especially important for e-Government applications, since these are mostly intended to be used be the “average end user”, who is not familiar with technological details.
The inexperienced users were neither able to judge the capabilities of MultiStates, nor to compare the information available to the depth of information of a desktop product.
Some participants had problems interpreting the meaning of system icons on the iPhone platform, since they had not used such a system before. However, not only icons were misinterpreted, but also users were unsure of how to interact with the touch-based device. Participants often asked for help on what they are able to do when performing tasks.
Although the overall acceptance of the interaction techniques was high (4.33/5.0) and all techniques were described as intuitive, users did not like the tap hold/tilt gesture and the tilt control. They found both to be too imprecise to work with the Health Summary.
Expert Group In contrast to the Comparison Group, the experts were able to provide more detailed feedback on usability problems and interaction methods.
The expert users showed a high acceptance rate (4.00/5.0) for the perspective wall and mentioned that the availability of context information was helpful when zooming in closely.
A key criterion of the expert user group was the ability to use the application single-handed. As a result, they preferred the utilization of tap hold/tilt interaction in combination with tilt interaction to quickly switch between scrolling and zooming. Unlike the Comparison Group, the experts described swipe and pinch interactions as precise but too slow to be useful. However, all interaction techniques were described as intuitive to use.
The results for the question whether MultiStates is a good combination of a complex desktop solution and a mobile application are good (4.33/5.0). Furthermore, the users stated that neither more nor less detailed data is required to be usable for them.
Comparison Finally, we looked at similarities and differences between both groups. The most important findings are as follows.
Looking at more general results, it turns out that all users agreed on the diversion of “slow but precise” input (swipe and pinch) and “fast but imprecise” (tilt and tap hold/tilt). Experts tended to prefer the faster input methods, whereas the inexperienced users liked the more precise techniques better.
It turned out that the Expert Group users performed timed tasks faster (Figure 3) and needed fewer corrections than participants of the Comparison Group (CG: 31, EG: 23 corrections – Figure 4).
Although the Comparison Group needed more corrections than the Expert Group, the number of corrections decreased constantly, except for two peaks in tasks four and six, from task three to task seven. Nevertheless the experts show a more constant number of corrections (based on the lower standard deviation).
Even though some users complained about reflections on the display while tilting the phone, the Expert Group and Comparison Group both did not think that tilting and shaking the phone distracted from working with the application. Especially when shaking the phone, users argued that they knew the result of the action (reset) and therefore did not need to see what happened on the display. Blind navigation feedback Both groups had to discover an alert or exception in the “Memory Usage” category. To achieve this, they were allowed to configure the preferences menu normally (e.g. activate the audio indicator and set the category filter to “Memory Usage”). As the next step, they had to zoom in and navigate through the databases having their eyes closed. Both Expert and Comparison Group found it neither easy nor difficult to solve this task (CG: 3.33/5.0 EG: 2.66/5.0) although the experts needed significantly less time to complete this task (see Figure 3).
Furthermore, both groups used the tap hold/tilt gesture to quickly zoom in, since it was difficult for them to coordinate two fingers on the phone without being able to see to perform a pinch gesture. Once they managed to zoom in close enough to make use of the audio focus, they just lifted their finger, which was used for tap-hold-zoom, and thus started to scroll through the Health Summary. NON-VISUAL EXLORATION: LESSONS LEARNED All users completed the non-visual exploration and discovery task successfully. However, they suggested improvements for the support of visually impaired people using the perspective wall. Using tap hold/tilt to zoom in and out turned out to be very easy for the users (in this case even for the Comparison Group). But the level of magnification is hidden from the user. As shown in Figure 2, four main levels of information depth are available. Using speech or different sounds as output, these levels may be announced to the user.
Furthermore, the borders of the perspective wall cannot be recognized by blind(-folded) users. They know when they discover a problem they adjusted the filter settings for, but they do not recognize when they reach the edges of the wall when they scroll by tilting the phone.
In addition to the acoustic representation of the current position within the perspective wall, an indicator for the initial position could enhance the feedback. As tilt control may be sensible for slight movements of the phone, it is important at least to know when the phone is held correctly in a way that no action results.
DISCUSSION Besides the characteristics of each group described above, it is clear to see that the overall error rate is not optimal, yet. None of the users complained about frustration during the use of MultiStates. Nonetheless, the Expert Group performed significantly better than the Comparison Group in terms of corrections and speed. Independently from the preferred interaction technique of each user, all techniques have been described as intuitive.
Additionally, the acceptance of the perspective wall is high among both groups (CG: 3.33/5.0, EG: 4.0/5.0). Nevertheless, some expert users pointed out that the screen becomes slightly cluttered with all the state information once they zoomed out completely. Creating more visible grids between each database and its states could lessen the cluttering effect.
Errors through accidently tapping the screen have not been counted since the cost of these errors (activating other features or performing different actions) is low. This relates to the definition of mostly tap-independent gestures. Nevertheless, errors in MultiStates sometimes forced users to repeat an action or to reset the view. The resulting actions have not been added to the number of user corrections.
In the end we were satisfied with the overall acceptance of MultiStates. Especially since none of the users had experience in using touch-based devices before and all of them were able to work with the system after a short demonstration of the interface and the available input techniques CONCLUSION As the introduction described, e-Government is highly dependent on a working IT infrastructure. MultiStates provides support for maintaining this infrastructure. Looking at the main aim of this work, the construction of a mobile and easy-to-use database performance monitoring solution, we can say, that the work was a success. The overall acceptance of it, especially within the Expert Group, was very good and no user failed to complete a certain task.
But the most interesting finding was that even through a simple acoustic indicator, non-visual exploration of an information space can be supported. Even though this is not sufficient to serve as a solution for visually impaired and blind users, it lays out the foundation for further work with touch-based devices.
A platform like the iPhone offers a considerable degree of freedom in creating user interfaces. In particular it offers to integrate the proven scalability of the perspective wall with application-specific set of gestures and allows the display of large datasets on a mobile platform. By offering scroll actions along both X- and Y-axis, we were able to display even more information in combination with geometric and semantic zoom. The flexibility of creating graphical user interface elements independently from hardware buttons and switches is a major advantage and may outweigh missing haptic feedback and lower hit accuracy to some degree.
FUTURE WORK Based on the evaluation with IT amateurs and experts we were able to discover leveraging points for the further improvement of the user experience. Most of these points indicate that a longer study with more and different evaluation participants would be helpful in order to improve the usable access of this application.
To get more detailed feedback on the MultiStates it would be feasible to have DBAs instead of amateurs and experts evaluate the system.
By providing more acoustic or tactile feedback, in particular a screen reader or screen magnifier, visually impaired and blind users can be further supported in using MultiStates.
By ensuring a high compliance to accessibility guidelines,
such as the Web Content Accessibility Guidelines (WCAG)
[
As mentioned above the use of a screen reader (e.g.
VoiceOver [
Once the enhancements mentioned previously have been incorporated into MultiStates, it would make sense to have blind users evaluate the system and to compare these results to the existing ones.
ACKNOWLEDGMENTS We thank IBM Deutschland Research & Development GmbH for supporting this work by making IBM DB2 Performance Expert and IBM Data Studio Administration Console available and for making the expert evaluation sessions possible.