Transcript 2.MakingDatabaseSyst..

Making Database Systems Usable
University of Michigan
H.V. Jagadish, Adriane Chapman, Aaron Elkiss,
Magesh Jayapandian, Yunyao Li, Arnab Nandi, Cong Yu
Mar 27, 2008
Christiano Santiago
1
Agenda
 Users Expectations
 The Persistence of Pain
 The Painless Future
 Conclusion
 Q&A
Mar 27, 2008
Christiano Santiago
2
Users Expectations
Database interactions
 Ability to query in a more sophisticated way,
using complex query semantics
 More precise and complete answers
 Expectation of structure in the result set
 Create and update databases
Mar 27, 2008
Christiano Santiago
3
The Persistence Of Pain
Mar 27, 2008
Christiano Santiago
4
Higher Level Presentation Data Model
Painful Relations
Mar 27, 2008
Christiano Santiago
5
Higher Level Presentation Data Model
Painful Relations
Simple concept versus Implementation
X
Flight info
Mar 27, 2008
Christiano Santiago
6
Higher Level Presentation Data Model
Painful Relations
What are the fligth departure times from Detroit to Beijing?
SELECT s.departure_time
FROM schedule AS s,
flight_info AS f, airports AS d, airports AS a
WHERE s.id = f.schedule_id
AND f.fid = d.id
AND d.city_name = “Beijing”
AND f.tid = a.id
AND a.city_name = “Detroit”
Mar 27, 2008
Christiano Santiago
7
Higher Level Presentation Data Model
Painful Relations
Problems
 Large number of table
 Poorly named entities
 Poorly named fields
 Joins are not intuitive
 Recursive self-joins
 SQL is too onerous
 Dependency on DBA (please don’t leave!)
Mar 27, 2008
Christiano Santiago
8
Multiple Ways to Access Data
Painful Options
Mar 27, 2008
Christiano Santiago
9
Multiple Ways to Access Data
Painful Options
Too much functionality and too many options
X
Mar 27, 2008
Christiano Santiago
10
Multiple Ways to Access Data
Painful Options
Barry Schwartz, The tyranny of choice.
Scientific American, April 2004, pp. 71-75
Mar 27, 2008
Christiano Santiago
11
Multiple Ways to Access Data
Painful Options
Perspectives
 Design systems for customized value
 Care about how well uses can get their job
done
 Forms-based interfaces have been the
primary means to interact with databases
 Provide simple querying tools for novice
users and powerful tools for experts
(customization)
Mar 27, 2008
Christiano Santiago
12
Precision and Recall
Unexpected Pain
Mar 27, 2008
Christiano Santiago
13
Precision and Recall
Unexpected Pain
X
Mar 27, 2008
Christiano Santiago
14
Precision and Recall
Unexpected Pain
Search engines error acceptance
 Expectations are set correctly
 Usually top few results do contain something
relevant
 If no relevant results were found, search it
again
 The web is so huge that no one knows
exactly what is out there
Mar 27, 2008
Christiano Santiago
15
Mar 27, 2008
Christiano Santiago
16
Precision and Recall
Unexpected Pain
Unable to query
 Systems may impede users from querying the way they want
 Scenario: world traveler with infinite flexibility and many
destinations to visit
X
Mar 27, 2008
Christiano Santiago
17
Precision and Recall
Unexpected Pain
Unable to query
 User’s versus form designer’s mental model of the data.
 Users should be worried about how. They just need the
what.
 Challenge #1: it’s not easy to provide a straightforward
and comprehensive way of specifying what a user wants
when a query requires aggregation or multiple complex
joins.
 Challenge #2: performance, security access level, data or
program errors may prevent users from running a query.
Mar 27, 2008
Christiano Santiago
18
Precision and Recall
Unexpected Pain
Unexpected results
 (Unexpected results) + (no explanation) = frustration
Example 1
• The client record shows he has 2 cars
but
• There is only 1 car related to this client in the
database.
# of cars was a field
in the Client table,
with no referential
Integrity with
CarOwner table
Mar 27, 2008
Christiano Santiago
19
Precision and Recall
Unexpected Pain
Unexpected results
 (Unexpected results) + (no explanation) = frustration
Example 2
User is looking for cheap flights provided by her
favorite airline:
• Los Angeles for $75
• Boston for $100
$400 was less
than half the
• San Francisco for $400 (?)
average price
for a ticket to
San Francisco
Mar 27, 2008
Christiano Santiago
20
Precision and Recall
Unexpected Pain
Unexpected results
 Empty & non-empty results
 Users may need to know where a result
comes from
 Users may need to know why a result is
returned
 System’s usability can be significantly
affected when no explanation can be given
Mar 27, 2008
Christiano Santiago
21
Errors In Query Construction
Unseen Pain
Mar 27, 2008
Christiano Santiago
22
Errors In Query Construction
Unseen Pain
Database Users versus Computer Scientists
X
Mar 27, 2008
Christiano Santiago
23
Errors In Query Construction
Unseen Pain
Mar 27, 2008
Christiano Santiago
24
Errors In Query Construction
Unseen Pain
Perspectives
 Query purpose: provide information that the
user doesn’t have
 How do I provide a clear and precise “from”
and “where” clause, if I don’t know the
possibilities?
 WYSIWYG
Mar 27, 2008
Christiano Santiago
25
Errors In Query Construction
Unseen Pain
WYISWYG
Assisted Querying using Instant-Response Interfaces
Arnab Nandi & H. V. Jagadish - University of Michigan
Mar 27, 2008
Christiano Santiago
26
Errors In Query Construction
Unseen Pain
WYISWYG
 Google Earth approach
 Can you image finding a location in a map
using coordinates instead of zoom and
rotation?
Mar 27, 2008
Christiano Santiago
27
Database Creation
Birthing pain
Mar 27, 2008
Christiano Santiago
28
Database Creation
Birthing pain
Corporate Market versus Average User
X
Mar 27, 2008
Christiano Santiago
29
Database Creation
Birthing pain
Perspectives
 Users want to store their data into databases
but this is not an easy task





Database schema
Table structure
Normalization
Relationships
Constraints
Mar 27, 2008
Christiano Santiago
30
Database Creation
Birthing pain
Perspectives
 An everyday database



No clear view of required database structure
No comprehensive design at starting point
Structure grows as time passes
 Structure grows incrementally
 Need to support both structured and
unstructured information
Mar 27, 2008
Christiano Santiago
31
The Painless Future
Mar 27, 2008
Christiano Santiago
32
The Painless Future
Geographic Model
 Joins between data sets using geographic or
spatial location as the basis
Mar 27, 2008
Christiano Santiago
33
The Painless Future
Network Model
 Graph or network representation of data.
Mar 27, 2008
Christiano Santiago
34
The Painless Future
Multidimensional Model
 Data as points in multi-dimensional space
Mar 27, 2008
Christiano Santiago
35
The Painless Future
Tabular Model
 People is used to seeing data represented in
simple two-dimensional tables
Mar 27, 2008
Christiano Santiago
36
The Painless Future
Presentation Data Models
Perspectives
 Data manipulation through presentation data model versus view
updates
 Consistency among view options: geographic and textual views
in synchrony
 Data provenance: “why” and “where” related to result sets
 Direct data manipulation


Point-and-click, drag-and-drop, etc.
New few number of algebra operators to support it
 Schema-later (unstructured) and heterogeneous database
design
Mar 27, 2008
Christiano Santiago
37
Conclusion
 Databases are extremely difficult for most
people to interact with.
 We must rethink the architecture of the
database systems as a whole.
 Presentation model as a distinct layer above
usual logical data model.
Mar 27, 2008
Christiano Santiago
38
Questions & Answers
Mar 27, 2008
Christiano Santiago
39