Transcript ppt

Face-off: Should we teach Synthesis
Algorithm and Related "Theoretical"
Topics in Database Courses?
Ivan Luković,
University of Novi Sad,
Faculty of Technical Sciences
&
University of Montenegro,
Faculty of Science, Podgorica
10th Workshop DAAD
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
Agenda
• Heritage and Alternatives in Databases
• Past and Present
• Issues and Experiences
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
2 / 25
Heritage and Alternatives in Databases
• A brief history of data models
– Many data models and paradigms in database (db)
design during the last four decades
– Implementation data models
• provided by commercial DBMSs
– Conceptual data models
• mostly used in the db design process
• to create conceptual db schema specifications
• that may be characterized as platform independent (PIM)
– Still "theoretical" data models
• provided by some DBMSs
• designated as very "new" and "promising", but
• a number of their commercial applications in industry practice
is still rather small
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
3 / 25
Heritage and Alternatives in Databases
• A brief history of data models
– Implementation data models
• the most favorable is still Relational Data Model (RDM)
– a long tradition and considerable popularity
– very extensive commercial usage and broad application
– we may believe that it will continue
» thanking to its sound theoretical, i.e. mathematical
background
– many logical problems, their solutions, and principles
concerning db design are formally defined at the level of RDM
– Conceptual data models
• the most favorable is still Entity-Relationship (ER) data
model
– also a long tradition
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
4 / 25
Heritage and Alternatives in Databases
• Design approaches
– A typical scenario of db schema design process
• (1) create the first an ER database schema and then
• (2) transform it into the relational database schema
– Many advantages
– A number of CASE tools providing such
transformation process
– It is based on applying well-known mapping rules
• with a common belief that such rules will guarantee the
satisfaction of the Third Normal Form condition (3NF) per se
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
5 / 25
Heritage and Alternatives in Databases
• Design approaches
– A typical scenario of db schema design process
• (1) create the first an ER database schema and then
• (2) transform it into the relational database schema
– Unfortunately, a reality is different
• many obstacles or pitfalls may appear on that road
– Serious disadvantages - lack of semantic problem
• some mapping rules may be characterized only as advisable
• the transformation process depends not only on the formal
mapping rules, but also on the problem domain semantics
• many examples in which the same structure of ER db
schema should not be transformed into the same relational
db schema structure
– due to the different semantics assigned to the ER structure
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
6 / 25
Heritage and Alternatives in Databases
• Design approaches
– A typical scenario of db schema design process
• (1) create the first an ER database schema and then
• (2) transform it into the relational database schema
– Main consequences
• a designer is forced to make some further corrections of a
relational db schema
– obtained by the transformation algorithm
• the source and target db schemas become unsynchronized
• the quality of such designed db schemas is highly dependant
of designers'
– theoretical knowledge and previous experience
– perception power and ability to recognize an influence of the
problem domain semantics onto the transformation process
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
7 / 25
Heritage and Alternatives in Databases
• Design approaches
– Many principles concerning db design are formally
defined at the level of RDM, but unfortunately
– Only some of them are widely exploited in practice
– An alternative approach
• db schema design based on well-known synthesis algorithm
– originally defined by [Beeri C., Bernstein P.A.; 1979]
• provides a powerful formalization of the relational db schema
design process, but
• suffers from a number of problems blocking its practical
usage
– still rarely exploited in practice
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
8 / 25
Heritage and Alternatives in Databases
• Our research results provide
– Raising the synthesis algorithm to the level of its
practical application
• in large db design projects
• in the scope of Model Driven Software Development (MDSD)
process
– Overcoming the problems of db schema design
based on ER data model and transformations into
RDM
• we have made necessary improvements of the synthesis
algorithm and
• embedded it into our MDSD tool, named IIS*Case
– presented two years ago
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
9 / 25
Agenda
• Heritage and Alternatives in Databases
• Past and Present
• Issues and Experiences
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
10 / 25
Past and Present
• Academy World
– Past (app. two decades before)
• theoretical and formal foundations of databases used to
teach in detail
– with considerably less problems felt by teachers
• students did not complain much about it
– they were to learn it, and they learned it
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
11 / 25
Past and Present
• Academy World
– Current state
• faculties are fighting for students
– surviving formula for both state and private faculties is
– STUDENTS + LESSONS = MONEY
• deans are frightened of loosing students
• a majority of students are frightened of theory, mathematical
formalisms, and reading books in general
– they believe that Internet is the only source they are supposed
to consult for everything in their life
» just a brief look at Wikipedia, Facebook / forums and
Google Services is enough for everything
– they are always complaining about theory and "deep"
formalisms, by default
• as a result, significantly reduced level of theoretical
knowledge is transferred to students, in average
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
12 / 25
Past and Present
• Research World
– Past (app. two or three decades before)
• theoretical and formal foundations of relational databases
used to develop intensively
• many logical problems in db design were resolved at the
level of RDM
– a heritage that must be tailored in any new data model being
developed
• in late 90's, theoretical and formal approaches in databases
were almost forgotten
– they were not "fashionable" enough for hot research
– Current state
• coming back to the roots, through XML data model
– the same fundamental principles and ideas evolved
• employing MDSD and DSL approaches is promising
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
13 / 25
Past and Present
• Software Industry World
– Past (app. two or three decades before)
• a considerable number of large and medium-size companies
– from various industry and government sectors
– having their own informatics departments
– with the main goal not only to support the maintenance of a
company IS, but to really perform its development
– large software development centers
» employed predominantly high (academic) educated people
• invested a lot of money in various training programs
– to raise the level of technological, as well as the methodological
knowledge necessary for IS development
• at the same time, a lot of rather small software companies
– specialized for IS development and outsourcing software
services
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
14 / 25
Past and Present
• Software Industry World
– Current state
• large and medium-size companies from various business
sectors mostly become just buyers of software systems
– many functions of the former informatics departments have
been outsourced
» reducing their responsibilities only to a minimum of
maintenance
» reducing the number and education profile of the staff
• sellers are predominantly large software companies
– offering their general purpose software systems that
» may be localized, parameterized or customized
» are usually independent of any particular DBMS
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
15 / 25
Agenda
• Heritage and Alternatives in Databases
• Past and Current State
• Issues and Experiences
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
16 / 25
Issues and Experiences
• Deans' messages
– Please, slow down with your theory and database
schemas, otherwise I will run out of all my students!
– You should realize that today's students are not the
same as you were!
– How do you imagine to earn your salary without
students?
– Of course, theory is very important. You should
preserve the quality of lecturing process for sure, but
be careful to provide a high ratio of passed students!
– Take an elective course and do there whatever you
want, but don't touch in basic, mandatory courses!
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
17 / 25
Issues and Experiences
• Students' messages
– Positive:
• Theory is really valuable to recognize all important issues in
databases.
– Pragmatic, funny, or possibly negative?
• Who even cares about theory in database design process? It
is completely unnecessary.
• Please, teach us only the practice. We need just a practice.
• We are always doing design in an intuitive and practical way.
• I know the process very well. Just simply draw up some
tables and link them by foreign keys. It is really easy, isn't it?
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
18 / 25
Issues and Experiences
• Colleagues' messages
– Average students fill great troubles in adopting the
theory and mathematical formalisms of RDM
• just like formalisms in other areas, such as languages and
compilers
– They even couldn't perceive the synthesis algorithm,
at all
– A common advise: remove the synthesis algorithm
and "high" theory from the database courses
• this is reported by many colleagues from various universities
– Alternatively, you may include such topics in the
elective advanced courses, but with a special care!
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
19 / 25
Issues and Experiences
• Messages from software sellers
– Suggestive, and usually accepted without any serious
analysis in a today's high-speed life
• What company is to develop its own database and IS? It is
time consuming and does not pay off, at all!
• Just buy our solution. It will resolve all your problems in the
best way and it is the cheapest solution for you!
– Many satisfied customers of our solution testify that for many
years.
• If you buy our solution, you may choose any DBMS you like.
You will not be dependant of any DBMS provider.
• We can efficiently and easily customize a solution for you
to ideally fit to your business!
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
20 / 25
Issues and Experiences
• Main issues
– Unfortunately, unforeseen and not seriously perceived
side effects and expenses
– DBMS independence:
• database schema will contain only tables, columns, primary
keys, and probably simple foreign key constraints
– nothing more than that
– a power of DBMS is exploited in a really small extent
• everything else is an enormously complex middle-layer
business logic deployed at application servers
– latter on, a customer has to pay a great amount of money for its
maintenance and trivial customizations, just to that seller
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
21 / 25
Issues and Experiences
• Main issues
– Unfortunately, unforeseen and not seriously perceived
side effects and expenses
– "Easy" customizations may result in:
• careless adding a mass of new tables and columns
• poorly controlled, chaotic and erroneous database solutions
– However, all these messages are fully pragmatic!
– Even, there is a new message from all sides:
• We are aware of future potential problems, but let us resolve
our urgent problems, ASAP!
• And finally, who cares about the far future.
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
22 / 25
Issues and Experiences
• Academic experiences from FTS
– currently, two courses
• Databases, 4+3, VII semester, mandatory for all CS students
– about 80 students per year
–  20% 9's and 10's on the scale of grades 5-10
– includes theory and formalisms (RDM, relational algebra,
functional dependencies and normal forms)
• Database Systems, 3+3, VIII semester, elective
– No. of students:
» 7 in the first running year, 42,8% of 10's
» 22 in the second running year, 41% of 10's
– includes the synthesis algorithm and related "heavy"
formalisms, but
– also includes a GUI tool that provide its practical application in
database design process
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
23 / 25
Issues and Experiences
• Academic experiences from FTS
– from this academic year, three courses
• Databases I, 4+4, V semester, mandatory for all CS & CE
students
– about 140 students per year
– basic theory and formalisms
• Databases II, 2+2, VII semester, mandatory for all CS
students
– about 80 students per year
– moderate theory and formalisms
• Database Systems, 3+3, VIII semester, elective
– No. of students:
» we may expect 20 students per year
– strong theory and formalisms, but also
– balance between theory and practice
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
24 / 25
Issues and Experiences
• We believe that
– a respectable academic education system
• preserving the balance between theory and practice
– combined with customized training programs
– a utilization of theoretical foundations combined with
MDSD and DSL approaches
• throughout the whole database development process
has a considerable potential to do much more
for software industry than it really does today
• Who is responsible for the far future and
strategy decisions?
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.
25 / 25
Face-off: Should we teach Synthesis
Algorithm and Related "Theoretical"
Topics in Database Courses?
Ivan Luković,
University of Novi Sad,
Faculty of Technical Sciences
&
University of Montenegro,
Faculty of Science, Podgorica
10th Workshop DAAD
JCSE W-2010 / Ivan Luković
Ivanjica, 6 – 10. 9. 2010.