37. Prague Computer Science Seminar


Databases have been around for more than 50 years. Codd’s relational data
model from 1970 decoupled the form of data from the physical storage of that
data and introduced two important formal query languages: relational calculus
and relational algebra. At that time, the research was focused on development
of
non-procedural query languages whose expressive power is closer to
computational
completeness. The SQL language has become an “intergalactic dataspeak”
supporting objects, XML data, and even recursion. A lot of effort was devoted
to
the design of database schemas including integrity constraints ensuring
database
consistency. We developed new approaches to database modelling both at the
conceptual and data levels based on a simple theory of types, and used typed
lambda calculus for querying a functional database. This approach we used later
in the world of data warehouses, XML databases, and others. Our research was
also focused on implementation techniques, e.g., how to store and query XML
data
in a column-oriented storage.

The difficulties in scaling up SQL databases for online, web-scale processing
have led to the creation of NoSQL systems in the 2000s. But key-value, column,
document, and graph databases mostly do not use database schemas, integrity
constraints, high-level query languages, and do not guarantee properties
typical
for transaction processing in traditional databases. Of course, this is
unsuitable for many applications. The next decade offers a NewSQL databases -
modern relational databases that seek to provide the same scalability as NoSQL,
but guarantee the same transactional support as traditional DBMSs. We will show
this diverse database world and some possibilities how to approach it within an
integrated database architecture.